Why is the UV/IR filter not incorporated in M8

[arthury]

Advertisement (gone after registration)

"Increasing the filter thickness would have led to a deterioration of the image quality, especially in the frame corners. The solution of utilizing UV/IR filters that are attached on the lenses does not reduce image quality since the filter is located outside of the optical system." --- Leica M8 FAQ

 

I am not quite understanding the last statement. The light path will ultimately have to pass thru the IR filter whether it is on top of the sensor or the lens.

 

Why would placing the filter on the sensor make it lose resolution while placing it on the lens will not?

 

I thot the issue was because there isn't enough physical space to incorporate a thicker IR-reducing glass on the sensor (like the DMR) ?

[jaapv]

In front of the lens the light rays enter from infinity down to 70 cm. That is a whole different ball game from the sensor- rear element distance of say 15 to 30 mm. Plus , the angle of incidence is a minor variable, more or less, on the front of the lens, making firmware cyan falloff correction doable. On the sensor, though, the angle of incidence varies with the aperture and focussing distance, making correction a software nightmare, especially as the camera does not know the aperture exactly and the focussing distance not at all. And then there is the problem that IR filters absorb (to thick for high resolution) or reflect the IR light. When it reflects, the IR light starts bouncing around from the sensor to the lens and anywhere, creating more undesirable optical effects.

[cldp]

Light will be traveling fairly straight through a filter when it is in front of the lens. With a sensor cover filter the light hitting the sensor, especially light in the corners with wide angle lenses, will be going through the glass obliquely. That will cause increased diffraction and loss of acutance.

[Guest guy_mancuso]

"Increasing the filter thickness would have led to a deterioration of the image quality, especially in the frame corners. The solution of utilizing UV/IR filters that are attached on the lenses does not reduce image quality since the filter is located outside of the optical system." --- Leica M8 FAQ

 

I am not quite understanding the last statement. The light path will ultimately have to pass thru the IR filter whether it is on top of the sensor or the lens.

 

Why would placing the filter on the sensor make it lose resolution while placing it on the lens will not?

 

I thot the issue was because there isn't enough physical space to incorporate a thicker IR-reducing glass on the sensor (like the DMR) ?

 

 

In simple terms on the lens the IR is blocked and does not hit the sensor

[rvaubel]

In front of the lens the light rays enter from infinity down to 70 cm. That is a whole different ball game from the sensor- rear element distance of say 15 to 30 mm.

 

Not true. It is the ANGLE of the light with respect to the surface of the filter that causes the cyan vignetting. It doesn't matter where the filter is located, front or back.

 

I figure the reason Leica did what they did (use a minimum cut filter...too minimum) was to avoid the excessive cyan vignetting they knew they would get on wide angle lenses if they used a stronger filter. They just hoped that no one would notice the magenta cast on highly reflective IR surfaces. However, it was a design tradeoff, not a plot or sign of incompetance.

 

What I do find intriging is the possibility of inserting the cut filter inside the lens, at a point that the rays are near normal(90 degrees) to the filter plane. If this could be done, the cyan vignetting effect could be eliminated. I think jaav has had this experience with a inserted filter on a superwide 9mm lens.

 

The possibility of Leica, or anyone else, designing an ultrawide specifically for the M8 could solve a lot of problems without the necessity of an external filter.

 

What do you think? Am I thinking right or just hallucanating?

 

Rex

[jaapv]

I don't think you read my post, Rex. The angle is determined by the distance between the source and the lens or the sensor and the lens. Light coming from infinity and entering the lens has a different angle from light traversing 20 mm to the sensor from the rear of that lens. The main argument in the full-frame vs crop discussion is exactly that angle. And the angle with which the light hits the sensor is identical to the angle that it exits the lens with.

Your point of inserting the filter in the lens is well taken. As I posted elsewhere, the 9.8 mm Tegea has the filter next to the diaphragm, inside the optical cell and exhibits no cyan vignetting.So no- you are not hallucinating.

[scott kirkpatrick]

Advertisement (gone after registration)

I don't think you read my post, Rex. The angle is determined by the distance between the source and the lens or the sensor and the lens. Light coming from infinity and entering the lens has a different angle from light traversing 20 mm to the sensor from the rear of that lens. The main argument in the full-frame vs crop discussion is exactly that angle. And the angle with which the light hits the sensor is identical to the angle that it exits the lens with.

Your point of inserting the filter in the lens is well taken. As I posted elsewhere, the 9.8 mm Tegea has the filter next to the diaphragm, inside the optical cell and exhibits no cyan vignetting.So no- you are not hallucinating.

 

I think Rex is right and you have a strange sense of geometry. The light entering the front of the lens and destined to end up at the edges of the image goes through the filter at the angle of view -- 45 degrees off axis for a 21mm equivalent lens, for example. The angle is the same whether the light source is the sun or just a few feet away. The angle at which it leaves the lens and strikes the sensor at the edge can sometimes be less, since the exit pupil -- the point in the lens at which the aperture appears to sit -- is often further from the image plane than the focal length. That's why IR filters are generally put behind a lens, where the angles are less dramatic.

 

scott

[sean_reid]

What I do find intriging is the possibility of inserting the cut filter inside the lens, at a point that the rays are near normal(90 degrees) to the filter plane. If this could be done, the cyan vignetting effect could be eliminated. I think jaav has had this experience with a inserted filter on a superwide 9mm lens.

 

What do you think? Am I thinking right or just hallucanating?

 

Rex

 

Hi Rex,

 

I'm told by an accomplished lens and filter designer that an internal filter of this type might work quite effectively.

 

Cheers,

 

Sean

[jaapv]

I think Rex is right and you have a strange sense of geometry. The light entering the front of the lens and destined to end up at the edges of the image goes through the filter at the angle of view -- 45 degrees off axis for a 21mm equivalent lens, for example. The angle is the same whether the light source is the sun or just a few feet away. The angle at which it leaves the lens and strikes the sensor at the edge can sometimes be less, since the exit pupil -- the point in the lens at which the aperture appears to sit -- is often further from the image plane than the focal length. That's why IR filters are generally put behind a lens, where the angles are less dramatic.

 

scott

 

I apologize. I was too simplistic.IR filters are put behind the lens on lenses with a long register, where your explanation often holds true. RF lenses have a short register, so it will often be just the opposite, depending on the exit pupil. For instance the 24 asph has a smaller exit angle, so it is as you say there. But with the Apo Summicron 90 it is just the other way around. Which makes a fillter at the back by no means a universal solution, only for certain lenses. The best solution would be an internal filter, if needed in an internal revolver, like for instance the 19 mm Elmarit R.

[barjohn]

How much clearance is there between the back of the lens and the front of the shutter? Could a filter be placed in front of the shutter but behind the lens (sort of like the SD 14)?

[Guest guy_mancuso]

The rear element is way to close to the sensor to even think about the back side , maybe on the inside rear element maybe the best choice.

[AlanG]

If I remember my optics correctly, a filter behind the lens will displace the focus a distance equal to the thickness of the filter. Each air to glass surface will defract the light rays first in one direction and then back in the other. It becomes a factor as part of the optical system.

 

Lenses that accept rear glass filters as part of their design require some kind of filter to be installed at all times for sharp results. I used to use rear filters pretty often with my view camera and I had to refocus when I put on a glass filter. (Gels were ok.)

 

So you probably don't want to use rear filters with a rangefinder system (unless you recalibrate the focusing mount.)

[arthury]

If I remember my optics correctly, a filter behind the lens will displace the focus a distance equal to the thickness of the filter. Each air to glass surface will defract the light rays first in one direction and then back in the other. It becomes a factor as part of the optical system.

 

Lenses that accept rear glass filters as part of their design require some kind of filter to be installed at all times for sharp results. I used to use rear filters pretty often with my view camera and I had to refocus when I put on a glass filter. (Gels were ok.)

 

So you probably don't want to use rear filters with a rangefinder system (unless you recalibrate the focusing mount.)

 

That can easily be done by the lens manufacturer if that's their design. I am not suggesting putting a filter behind the lens after designing the lens otherwise.

[AlanG]

That can easily be done by the lens manufacturer if that's their design. I am not suggesting putting a filter behind the lens after designing the lens otherwise.

 

Yes, new lenses will be a solution. But at that point why not incorporate a chip and transfer f stop and distance information too?

 

And of course new lenses could be designed incorporating a filter. But simply providing screw on filters seems to have been a challenge. They could also have designed new wide angle lenses that are more tele-centric and then they could have used a better filter over the sensor in the first place. But I thought the point of the M8 was to have compatibility with existing lenses.

 

So there seem to be trade-offs. Correct the existing lenses as well as possible. Or make new lenses with internal or rear filters that work better on the M8, or make an M9 with a better IR filter over the sensor and have some w/a lenses specifically optimized for that. (And correct for the old lenses as best as possible.) I am not a camera designer or an optical engineer. I have no idea what is the best approach or if it would be within the company's resources and budget.

 

I think with the M8, Leica had its work cut out just getting to this point. And Leica probably didn't want to mess with the overall system too much. (Compatibility and overall M-ness is a big part of the appeal.) For those who don't own any Leica lenses, it would have been more appealing if Leica had started from scratch and made a camera and lens system where each item was optimized as part of the overall design solution. However that was acheived. For instance the 28mm lens for the Hasselblad H3 was designed with digital correction in mind.

 

My guess would be that the M9 will have an IR filter over the sensor and some w/a lenses that are made for it. It seems like the easiest approach to me.

[lars_bergquist]

The idea of internal IR filtering in the lens is attractive, especially for extreme wide-angle lenses. It should be interchangeable however. Many such lenses for SLR cameras have had filter turrets, in which case we could have e.g. an UV/IR, one ND and one clear glass. Another solution would be a filter drawer.

 

Interference filters do of course reflect some light forward (hence the 'Eye of Mordor' red glare) but if any of this would happen to be re-reflected back, it would just encounter the filter again an probably never reach the sensor. Listen, Cosina! Hark, Leica!

 

The old man from the Age of Old Men

[Goldie]

Leica's statement is perfectly correct if a little succinct.

My understanding is that for the sensor cover glass to have done the job of IR filtering, it would have had to be considerably thicker. The space behind the lense is already very small in M cameras. More importantly in the M system light rays at a high image height strike the image plane at a relatively shallow oblique angle, this is not a problem with film since it has no 'thickness'. However the cover glass of the sensor has to behave like any glass and have refraction properties at both entrance and exit surfaces, the thicker the glass the stronger the redirection of the light rays and hence the stronger the effect on image quality.

A filter in front of the lense is literally outside the optical system; light has not even struck the first lense surface!

Filters at the back of the lense would have been a nightmare to produce, and have been inside the optical system.

Filters inside the lenses would have been useless since you could not use the lenses with film M cameras - since IR filters cause colour changes.

Filter turrets are impossible in M lenses Lars, where's the space? Never mind the cost and the re-design.

[scott kirkpatrick]

This subject has been worked over pretty hard, but it is an interesting set of tradeoffs, which were undoubtedly made under time and cost pressures that forced a feasible, not an optimal set of decisions. Once the decision was made not to use a dichroic IR filter perhaps 100 microns thick over the sensor as was done in the DMR (because it would require firmware corrections for the red vignetting which would result with wide angle lenses) the decisions got steadily tougher -- use absorptive material, but discover that the material which was then available gave too weak a filtration, leading to a dichroic filter in front of the lens. This requires an even stronger correction in firmware than would have been required with the filter at the sensor, but at least it is dependent almost entirely on the lens focal length and perhaps easier to compute. I don't think when Leica started down this path they realized that they would end up back where they started.

 

Probably the least demanding position for an angle sensitive filter is inside the lens, between the exit and entrance pupils, but even at such a location (if it is even possible in a RF lens) there will be angle dependence. Some earlier poster said "put it inside, where the light rays are all going straight," or words to that effect, but it doesn't work that way. (If all the light rays went straight inside the lens, how would they know which way to bend when it's time to exit...?) Attached is a ray tracing picture of what happens, borrowed from an Erwin Puts article. You can guess the entrance and exit pupil location by extending the rays entering or exiting until they meet, but the actual rays between the first and last elements bend by various amounts, and on average go through smaller angles because the lens is stretched out beyond its effective focal length.

 

scott

rays.pdf

[artur5]

Reading the entire (and instructive) thread it seems clear that there's not a perfect solution for the I.R. issue -either with with filters in front, behind or inside the lens.

The answer is much more obvious ( not easy, or even feasible nowadays). Just a sensor without ( or very little ) sensibility to I.R., as simple as that.

( Some of you will say that's already invented and it's called "film".. I know... )

I realize Leica hasn't neither the financial resources nor the tech. capability to do that. This is a task for Fuji, Sony, Canon, or Kodak..If any of them were really interested I think they could do it.

Until then, I'm afraid we're stuck with filters and/or cropped sensors on digital rangefinders

[AlanG]

...The answer is much more obvious ( not easy, or even feasible nowadays). Just a sensor without ( or very little ) sensibility to I.R., as simple as that...

...This is a task for Fuji, Sony, Canon, or Kodak..If any of them were really interested I think they could do it. Until then, I'm afraid we're stuck with filters and/or cropped sensors on digital rangefinders

 

Or a sensor that won't have cyan drift with w/a non-retrofocus lenses even though it somehow filters out IR radiation without the use of separate IR blocking lens filters.

 

The sensor technology you are asking for seems to exist.

 

This has been posted before but I suspect that not all have seen it. I'm guessing that Canon has some kind of unique sensor, microlens or filter technology and that may be the reason that they have few issues with their full frame sensors. (Canon FF sensors don't need lens optimization as did the Kodak 14n for instance.)

 

This person used a Voigtlander 15mm lens on his 5D by raising the mirror and then having a little pad on the top of the internal lens protrusion for the mirror to bounce into as he shot photos. The results showed a little vignetting that may just be characteristic of the lens, but otherwise worked well.

 

Voigtlander 15mm v Sigma 12-24mm: Intro

 

This article was posted last year yet I haven't seen any other similar tests or reports from others trying this. (If the Canon cameras had a mechanical mirror lock-up, I'm sure more would try it.) The 1DMkIII with its live view might be a good model to use it on. He did not test specifically for cyan drift but I don't see any in the one full area shot that is posted.

 

It made me wonder...

[Advertisement]

Hallo,

Look here: Leica M

[rvaubel]

 

Probably the least demanding position for an angle sensitive filter is inside the lens, between the exit and entrance pupils, but even at such a location (if it is even possible in a RF lens) there will be angle dependence. Some earlier poster said "put it inside, where the light rays are all going straight," or words to that effect, but it doesn't work that way. (If all the light rays went straight inside the lens, how would they know which way to bend when it's time to exit...?) Attached is a ray tracing picture of what happens, borrowed from an Erwin Puts article. You can guess the entrance and exit pupil location by extending the rays entering or exiting until they meet, but the actual rays between the first and last elements bend by various amounts, and on average go through smaller angles because the lens is stretched out beyond its effective focal length.

 

scott

 

In your 21mm Elmarit example the preferred position for a filter seems to be between the 2nd and 3rd element. The off axis rays seem to be most normal at that position. Also note, from scott's example that a cut filter in front of the lens seems to be the worst position for this particular design.

 

It seems to me that Leica would find a market for an M8 specific, fast ultrawide that included an internal filter. I would buy a 25mm F2 or F1.4 internally filtered crop factor lens in a heartbeat. By reducing the coverage requirements to the smaller M8 sensor, it would be a lot easier to design a fast wide that is not too big. Of course the film quys would be up in arms. But really a ultrafast wide wouldn't really be of much interest to the full frame community, the need is for the M8.

 

Rex