What is the problem with IR filters?

[sean_reid]

Advertisement (gone after registration)

No, it's real. See the "cut filter" thread for examples at 21 mm. This is the reason that Leica asked Kodak to put the weaker (but less angle-dependent) IR filter into the KAF 10500 imager that is used in the M8. They hoped that an absorbtive filter, which is less angle sensitive, would be easier to deal with for all kinds of lenses (not just coded ones), than an interference filter as is used in the similar chip that is in the DMR.

 

The dichroic IR filter seems to work just fine in cleaning up the IR false color problem for lenses of 28mm and longer focal length. For wider angles you need some post processing action to reduce the cyan shift around the edges. The effect is magnified for the ultra-wides, compared to what we would have seen if the dichroic filter had been on the sensor, because these are retrofocus designs. The range of angles at which light strikes the sensor is less than the range of angles at which light enters the front of the lens.

 

scott

 

Hi Scott,

 

The filter may perform well with lenses a bit wider than 28 mm although the cyan is clearly present in Guy's 21 mm samples. I'll be doing some tests to see how wide one can go before hitting the cyan corners. And, of course, the filters Guy and I are using are not necessarily the same as the filters that Leica plan to provide so we don't know for sure yet how the performance of the former is related to that of the latter. In any case, I want to look at how the 24/25s do with this filter.

 

The following is just a hunch, *not fact* but I think that the need to correct for these cyan corners (where red is accidentally being filtered in addition to IR) is the reason that Leica wants to see these filters on coded lenses so that the camera can correct the shift. If that is correct, the coding will be significant for ultra-wide lenses but perhaps not for wide and longer lenses. It certainly has not been necessarily for the 28s I've been using on the M8.

 

Cheers,

 

Sean

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[plasticman]

Mani/plasticman,

 

the Cosina/Voigtländer 15 mm lens is like most other lenses of that and similar focal length: there is no filter thread. The bulging front lens is protected by a built-in shade of the now traditional 'petal shape'. In any case, tests with that lens that I have seen have been very good. I am beginning to suspect that the IR false colour problem too will be fairly adequately dealt with by firmware, i.e. profiles. It seems that already, JPEG profiles do a better job than those used with DNG. The remaining problem will probably be with incandescent lighting only, and I'm already using correction filters there!

--A 15 mm 1:4.5 lens is not exactly what you use with incandescent, and neither really are a 21 or 24 mm 1:2.8. One single 46 mm filter will take care of my 28 mm Summicron and 35 and 50 mm Summiluxes.

--On the other hand, when I get my M8 I may well buy a deep-red IR filter, set the camera to black-and white and go out run some experiments ...

 

Tack Lars! So many extra issues that i hadn't even thought about prior to the emergence of the filter 'solution'.

The 15 or the 12 i thought of as capturing what me and my brother tend to call the 'NatGeo-look': that is, you can get the entire Mongolian Yurt in one shot. ;-) But as i understand it, the IR problem isn't restricted to indoor lighting - so b&w may be the only real 'solution' for the camera after all...

[fotografr]

The dichroic IR filter seems to work just fine in cleaning up the IR false color problem for lenses of 28mm and longer focal length.

 

scott

 

Scott--When referring to 28mm, do you mean the actual focal length or the effective focal length including the crop factor?

[Guest guy_mancuso]

Thanks Scott yes in my testing of the filters the Zeiss 21 i did get cyan corners on all sides. I am hoping the 24mm that effect will be a lot less , just waiting on the filter to arrive to try that one. But the 21mm focal length and i will assume the leica 21mm also will have this effect. I did a quick fix of it with color dodge with a magenta color and painted over the area and with care it can be done successful but a more automated way would be better. If you read the Cut filter theread you will see some of the discoveries of some informal testing with and without the IR filters.

[davidada]

I wonder if a graduated IR filter similar to the one contax supplied with the Hologon would be a solution on the 21mm and wider?

[dhoelscher]

I wonder if a graduated IR filter similar to the one contax supplied with the Hologon would be a solution on the 21mm and wider?

 

It seems to me that the standard IR/UV cut filters from B+W are rare enough, much less a graduated one ....

 

So, I'll bet Leica is preparing firmware changes that will allow the JPG and the DNG files with wide (>28mm) focal lengths with the IR filters to have the cyan corners automagically removed with in-camera processing. This is why they stated that their upcoming solution was only for Leica lenses with 6-bit coding (so that the camera would know the focal length).

 

DH

[dhoelscher]

Advertisement (gone after registration)

Of course, those of us with non-Leica lenses wider than 28mm or so will need our own work-arounds independent of Leica ....

 

DH

[sean_reid]

It seems to me that the standard IR/UV cut filters from B+W are rare enough, much less a graduated one ....

 

So, I'll bet Leica is preparing firmware changes that will allow the JPG and the DNG files with wide (>28mm) focal lengths with the IR filters to have the cyan corners automagically removed with in-camera processing. This is why they stated that their upcoming solution was only for Leica lenses with 6-bit coding (so that the camera would know the focal length).

 

DH

 

As you probably know from my post above, that's what I suspect as well. I need to test to see if the max is 28, 25 or 24. That's an open question right now.

 

Cheers,

 

Sean

[sean_reid]

Of course, those of us with non-Leica lenses wider than 28mm or so will need our own work-arounds independent of Leica ....

 

DH

 

....unless Leica allows us to manually specify focal length in the menu. Now, of course, their corrections would be optimized for their own lenses but those corrections might also work for other lenses. Again, we shouldn't assume 28 as the widest until we have evidence. I will try to test the 25 today...don't have the filter yet that I need for the 24. Scott has seen my 28 pictures (37 mm EFOV) and they were fine in that respect. 24/25 is an open question.

 

Cheers,

 

Sean

[Guest guy_mancuso]

You know that has come up several times on this board about the menu option of selecting your wide angles through the menu and i am really starting to like that idea because like Sean said the canned profile may work fine with off brand lenses like Zeiss and C/V which a lot of folks already have in the kits. It woud be nice to have both setups the coded one that automatically puts the fix in for the wides but the option also to apply that through the menu. That to me is a very nice working solution. I wish a I had a 28mm right now to test that but all i have is a 24 and 21 and no filter big enough for the 24mm. i am thinking the 24 will be effected but it maybe minimal which i could live with until a fix is in

[rosuna]

I suppose Leica's IR filter will be softer than a normal IR filter. It only must to filter some of the IR light because the sensor will filter the remaining IR light. The task will be shared between two IR filters, one over the CCD and another over the lens.

 

THis "soft" IR filter for Leica lenses would be less problematic than normal IR filters. Am I wrong?

[Guest guy_mancuso]

Hard to say Ruben but you have a point. I don't know if the IR filters are taking too much IR out , certainly there getting the job done to block the IR but what we don't know is it going to far with the combination that is already in place on the sensor. It's a great question

[peterv]

gumshoecamus,

thanks for the reply.

 

Still, my basic question is:

 

What is the best solution for the IR-problem, if we want to maintain or improve present image quality?

 

Peter

[sean_reid]

You know that has come up several times on this board about the menu option of selecting your wide angles through the menu and i am really starting to like that idea because like Sean said the canned profile may work fine with off brand lenses like Zeiss and C/V which a lot of folks already have in the kits. It woud be nice to have both setups the coded one that automatically puts the fix in for the wides but the option also to apply that through the menu. That to me is a very nice working solution. I wish a I had a 28mm right now to test that but all i have is a 24 and 21 and no filter big enough for the 24mm. i am thinking the 24 will be effected but it maybe minimal which i could live with until a fix is in

 

Hi Guy,

 

As you know, I've given the 28 Summicron plus 486 a workout already. I'll try to do some tests on the Zeiss 25 with 486 later today. Need a 55 to test the Leica 21 and 24.

 

Cheers,

 

Sean

[sean_reid]

I suppose Leica's IR filter will be softer than a normal IR filter. It only must to filter some of the IR light because the sensor will filter the remaining IR light. The task will be shared between two IR sensors, over the CCD and over the lens.

 

THis "soft" IR filter for Leica lenses would be less problematic than normal IR filters. Am I wrong?

 

Hi Ruben,

 

It's an interesting thought but I think that we need to test these to really know the answers.

 

Cheers,

 

Sean

[sean_reid]

 

What is the best solution for the IR-problem, if we want to maintain or improve present image quality?

 

Peter

 

 

Right now, the 486 filters are the best solution I know of. That's why I'm using them.

 

Cheers,

 

Sean

[Guest guy_mancuso]

Hi Guy,

 

As you know, I've given the 28 Summicron plus 486 a workout already. I'll try to do some tests on the Zeiss 25 with 486 later today. Need a 55 to test the Leica 21 and 24.

 

Cheers,

 

Sean

 

 

That be great Sean as I am stuck in the mud without a bigger filter for my 24mm. I ordered so many filters I think I could start a store, I lost count. LOL

 

Love to see if the 21mm is better or worse than the Zeiss. If you want i could send you a raw Zeiss 21mm for you to compare. Just let me know i have that street one and also the group if this helps at all

[scott kirkpatrick]

Scott--When referring to 28mm, do you mean the actual focal length or the effective focal length including the crop factor?

 

Actual. The new 28/2.8 released at about the same time as the M8 got a real workout in Sean Reid's wedding pictures, posted last night. He used a front dichroic filter, from B&W, I think. As he points out, there may be additional things that Leica can do to make a front filter have good anti-reflective coating, or other tweaks to match with the lens for which it is sold, but that certainly could make a $80 fix into something much more expensive.

 

However, the correction for the fact that a dichroic filter cuts into the visible frequencies well off axis can be done by strengthening the observed RGB data to compensate. It's just like the vignetting correction that the M8 does so well already for coded lenses. The correction for an unknown amount of IR coming into all three channels cannot be done accurately based on the observed RGB data. You simply don't know what is out there to correct.

 

Notice that rangefinder lenses from 28mm up are pretty much symmetric designs. I don't know about the 24 and 25 mm examples but I assume that the Tri-Elmar angle, the 21mm and the exotic 12, 15, 19 (I have one to try) are retrofocus designs. For a symmetric lens, the "cyan vignetting" is the same whether you put the filter in back or in front. For a retrofocus lens, it gets to be much better to put it in back, over the sensor.

 

BTW, I got much of my information on this from the well-known posts of Joseph S. Wisniewsky, which are mostly in the dpreview Leica forum, plus a careful reading of the Kodak chip spec, which has probably been sitting, waiting for us to notice it for a year now. Joe has a modest proposal of how a very thin glass dichroic filter could be glued over the existing sensor, and even made a stab at a cost estimate on this. This, plus firmware for the coded lens cyan vignetting, and accessible parameters to let users deal with cyan vignetting on non-coded lenses, would be a pretty nice rescue from this problem, if it can really be done.

 

 

scott

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[scott kirkpatrick]

....unless Leica allows us to manually specify focal length in the menu. Now, of course, their corrections would be optimized for their own lenses but those corrections might also work for other lenses.

 

The starting point with a Zeiss or CV wideangle would be to paint black and white dots on it to fool the camera into thinking it was the closest Leica lens in characteristics. But better would be to allow us to enter the exit pupil distance (not the focal length, but the position of the point from which the light rays appear to come out the back of the lens -- this determines the angles that the light rays make at the sensor), then I suspect that there is one de-vignetting algorithm that will fit 'em all.

 

some problems -- what keyboard do we type on to enter this number? I guess it can be done with up and down arrows. Who knows the exit pupil of my, say Canon 19mm lens? I guess a forum research project could whip those up in a few days, but it has to be done.

 

scott

[ho_co]

The dichroic filters are recommended only up to an angle of incidence of 30° according to the Edmund Optical site. That would mean that on the M8, the 28mm is the widest lens to be usable without some cyan in the corners.

 

Consider this, which I think might be a better solution: Go with IR-absorptive filters over the lenses instead. That would mean some loss in visible light transmission, but would mean that we would have the same color with all lenses. (Leica's lenses are among the most color-similar across the line on the market.) If Leica took that approach, we could then go with a single menu setting to correct the color shift.

 

I've been a bit more boringly explicit at http://www.leica-camera-user.com/digital-forum/9074-cut-filter-must-see-12.html#post93254.

 

--HC