Survey: Your opinion about the new LEICA M MONOCHROM

[pico]

Advertisement (gone after registration)

[...]Regarding IR filtering. It is my understanding that most sensors use dichroic filters that reflect unwanted IR but that this wouldn't work with the M9 due to the requirement of the M's shorter lenses.

 

I do not understand that part. The disadvantage of a dichroic filter applies largely to lenses which project steep angles of light to the sensor so that we find edge, fringe discoloration. IOW, the coating over areas receiving steep angled light is somewhat defeated. In an MM this could create some loss of resolution in the areas effected because allowing long (low frequency) light to be sensed.

[Advertisement]

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

[AlanG]

I do not understand that part. The disadvantage of a dichroic filter applies largely to lenses which project steep angles of light to the sensor so that we find edge, fringe discoloration. IOW, the coating over areas receiving steep angled light is somewhat defeated. In an MM this could create some loss of resolution in the areas effected because allowing long (low frequency) light to be sensed.

 

I didn't say it would work in the MM either. Just that they might need a stronger version of the dye based IR filter. Purely speculative and this diverges from the matter.

 

I just want to make clear that I have absolutely no idea what is going on in either sensor and their associated filters, processing etc. I just question the statement about the removal of the Bayer filter "alone" giving just a one stop change assuming everything else is equal.

 

My point being that I doubt that everything else is truly equal. That is all.

[brianv]

Looking at the July 26, 2007 Rev 5.0 MTD/PS-0962 Data Sheet for the M8 sensor, the KAF-10500, the Blue pixels are at 10% QE at 550nm, which is on the far side of the Green peak. Green extends well into Red, 10% QE is at 630nm.

 

I love the colors from my M8, hard to argue with success.

[AlanG]

Looking at the July 26, 2007 Rev 5.0 MTD/PS-0962 Data Sheet for the M8 sensor, the KAF-10500, the Blue pixels are at 10% QE at 550nm, which is on the far side of the Green peak. Green extends well into Red, 10% QE is at 630nm.

 

I love the colors from my M8, hard to argue with success.

 

Yes and those filters probably absorb two-three stops of light.

 

The sample below shows the split channels. One clearly needs several stops of separation between channels in order for color to reproduce properly. If you don't get this separation when shooting through the filters how will this be achieved? And this sample is just from a paper target that is printed with inks and is not close to the range of colors in typical scenes.

 

Just to take yellow as an example. If the red and green channels need to be white or nearly white for a bright yellow, then the blue filter on the sensor better block most of the red and green light or you won't have any way to reproduce a vivid saturated yellow. (Keep in mind these are impure inks in the test chart.)

 

It would be great if color separation can be accomplished with weaker filters but I think that is not realistic.

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!

[brianv]

The spectral transmission of the filters is given in the datasheet, no need to guess. I did not guess or speculate, I read the datasheet.

 

The traditional color separation filters are Wratten 25, 47b, and 58. "Kodak Wratten Filters, Scientific and Technical Databook B-3, Eastman Kodak company 1965". Looking at the data sheets for these filters, they do not overlap past the 10% mark, the color filters used digital sensors have a lot of overlap. They simply do not absorb as much light as the traditional tri-color separation filters used in Technicolor and other color-separation systems. The technicolor process did not interpolate.

 

So- either you announce that all digital cameras simply do not reproduce colors accurately enough for your taste, go back to film, or have a company fabricate a sensor to your design. The last one- expensive. been there, done that.

 

Good thing the M9 Monochrome is finally coming out. We can avoid these discussions. And I had color artifacts in my M9 shots yesterday, using a 75 year old Zeiss Sonnar wide-open at F1.5. And at F4- ouch, too sharp for mosaic filters.

[AlanG]

The spectral transmission of the filters is given in the datasheet, no need to guess. I did not guess or speculate, I read the datasheet.

 

 

How many stops of light do they absorb? Do you know that it only takes a pretty pale color filter to absorb about a stop of light? And filters that pale won't be suitable for color separation. Again, if the blue filter cannot block most of the red and green light, how will the system be able to reproduce yellow? I can't figure out any way. Can you?

[trond]

Advertisement (gone after registration)

The color response of the M9 can be found here:

 

DxOMark - Leica M9

 

Compared to "clear" pixels, the Green and Blue is slightly less than half a stop down, the red response is one stop down.

 

This is in perfect alignment with the M M begin one stop more sensitive.

 

The exact number is 1.17 stops more sensitive (Square root of (1/1 + 1/1 + 1/(1.15)^2 +1/(2.56)^2)) for the G + G + B + R pixels.

 

If you take into account the IR filter of the M M, this is probably very close to 1 stop.

 

Best regards

 

Trond

[brianv]

For the KAF-10500, The QE numbers are given with the BS-7 0.5mm cover glass in place, which absorbs IR. The IR absorbing filter reduces QE by "about" 30%, looking at the IR sensitivity. The QE of Kodak's 6.8uM detectors was published at 85% for a detector without infrared absorbing cover glass and without mosaic filter, that number may be improved for the KAF-10500, but is not available.

 

The RGB QE is given as is 21%, 40%, and 32% respectively. Figure 85% as a base for a Visible+Infrared sensitive detector, and "about" 60% QE for visible-only allowing for 30% sensitivity in IR.

 

So with a QE of 40%, the green absorbs less than 1-stop, the blue absorbs about a stop, and the Red absorbs about 1.5 stops. And all together- the bayer Pattern Mosaic filter absorbs about 1 stop in the visible band after the BS-7 cover glass is in place.

 

Stupid BS-7 cover glass. Needs to go. Found my pictures with a full-spectrum camera and Magenta filter, color is just like Infrared Ektachrome E3.

 

I have the transmission curves for the CC 50Y, CC 50M, and CC 50C. I believe the tricolor filters that I mentioned have the spectral characteristics that you are thinking of. They are steep drop-offs with very little overlap, true separation. The mosaic filters of digital cameras are not the same as the color separation filters.

 

At this point: an experiment using a Monochrome M9 and a separation filter wheel vs an M9 color would be interesting.

 

If I get the M9 Monochrome, I just happen to have the color separation filter wheel. and an M9c.

 

Another reason to buy the M9m for me. I need a new monochrome camera to go with that color separation filter wheel.

[pico]

How many stops of light do they absorb? Do you know that it only takes a pretty pale color filter to absorb about a stop of light? And filters that pale won't be suitable for color separation. Again, if the blue filter cannot block most of the red and green light, how will the system be able to reproduce yellow? I can't figure out any way. Can you?

 

If you block red and green, you cannot get yellow in the RGB system, and if 'most' is blocked, then some remains and the human eye is most responsive to yellow, then green. It takes less intensity to visualize these colors.

[brianv]

And with all of this discussion about Color Mosaic Filters, the Poll numbers for "Monochrome M9- Yes, Perfect camera for me" just moved past 13%, up from 12.5% before this discussion started. Probably the people that don't want to think about absorbng colors. I doubt it was all the talk about Quantum Efficiency.

[AlanG]

If you block red and green, you cannot get yellow in the RGB system, and if 'most' is blocked, then some remains and the human eye is most responsive to yellow, then green. It takes less intensity to visualize these colors.

 

??? If the blue channel does not block enough red and green it will start to kill the saturation of yellow which is composed of red and green in the the other channels. You want the blue channel as close as possible to zero value for maximum yellow saturation. The same logic applies to all the other colors since the system has to simulate them via red, green, and blue values. I don't know where the acceptable limit of slope and overlap is with sep filters but one stop of filtering does not sound like nearly enough to me. If the rest of you think that can work, that is ok by me.

[brianv]

Rather than all of your thought experiments and speculation, download the spec sheets and look at the spectral response yourself. Then look at a color image made using the detectors with these color filters and tell us that we've all been fooled for these past 20 years of using Bayer Pattern Mosaic filters.

 

You lose 1 F-Stop with the color filters. That's the spec. Thought experiments and speculations are a simple waste of time when the facts and results are available at your fingertips.

 

But all of this was good for Monochrome sales, as evidenced in the Polls.

 

I have to believe that more than one employee of Truesense involved with the monochrome KAF-18500 is reading this thread.

 

Thankyou. I've waited a long time for a replacement for my DCS200.

[AlanG]

 

The RGB QE is given as is 21%, 40%, and 32% respectively.

 

Correct me if I'm wrong but wouldn't 21% be more than two stops, 40% be 1 1/3rd stops and 32 % be 1 2/3rd stops more or less?

[brianv]

.

 

The RGB QE is given as is 21%, 40%, and 32% respectively. Figure 85% as a base for a Visible+Infrared sensitive detector, and "about" 60% QE for visible-only allowing for 30% sensitivity in IR..

 

 

The Infrared Absorbing Glass brings the QE down about 30% of it's peak. I put that into the equation. Based on an 85% QE for visible+infrared, put the IR absorbing glass over it, get about 60% QE for a visible only, monochrome camera. I put those numbers in. So the Green pixels are at 40%QE compared with a 60%QE for the monochrome visible-only pixel.

 

So the green pixels are down about 2/3 a stop from a monochrome pixel. 1/2 of the pixels are green, makes up for the bigger loss in red. 60% for monochrome visible dropped to 40% for green. Blue is down 1 stop, 60% to 33%. the Red takes a dive, 60% to 21%. Poor red, a Victim of the IR cut filter. Stupid IR cut filter.

 

Just curious- how many here have actually used cameras and scanners based on color-separation filter wheels? I have a scanner and film recorder both sitting in boxes in the basement that use them. Before those, used a Matrix camera and Dicomed. This isn't Rocket Science. There were two-filter color systems over 100 years ago.

 

Dichroic filters: I have a DCS200c that uses one over the detector. It just does not cut out as much IR as required for color photography, users would stack Two IR Cut filters on the Lens to get rid of the excess IR. IR cut Dichroic filters are "Hot Mirrors", used to be labeled such. They reflect deep red as well as IR. With the Dichroic filter over the detector, this produces a reflection into the lens. Not a good thing. It was fun to watch the cats chase the reflections of the Hot Mirror filters. But that was about it.

[mjh]

??? If the blue channel does not block enough red and green it will start to kill the saturation of yellow which is composed of red and green in the the other channels. You want the blue channel as close as possible to zero value for maximum yellow saturation.

You would if there was no colour space transformation from the device-dependent colour space of the sensor (which is partly determined by the colour filter array) into a device-independent colour space such as sRGB or Adobe RGB. But of course there is. The RGB values in the image must not be confused with the output of the red, green, and blue sensor pixels which may be quite different.

[jaapv]

For the KAF-10500, The QE numbers are given with the BS-7 0.5mm cover glass in place, which absorbs IR. The IR absorbing filter reduces QE by "about" 30%, looking at the IR sensitivity.

Iirc, the IR filter on the M9 was specified as 0.8 mm at introduction.

[mjh]

Correct me if I'm wrong but wouldn't 21% be more than two stops, 40% be 1 1/3rd stops and 32 % be 1 2/3rd stops more or less?

These percentages specify the quantum efficiency for the complete sensor assembly, i.e. including the IR-absorbing cover glass, the colour filter array, and the quantum efficiency of the sensor chip itself. If the QE for red is 21 percent then it doesn’t follow that the red filters absorbs 79 percent of all incoming light (unless the cover glass was actually clear and the chip had a QE of 100 percent, but then we know both to be false).

 

Btw, these figures are for the KAF-10500, i.e. the sensor of the M8. The figures for the actual sensor in the M9 are slightly different, due to changes in the colour filter array and a thicker cover glass (0.8 mm rather than the 0.5 mm of the M8) with increased IR absorption.

[brianv]

Iirc, the IR filter on the M9 was specified as 0.8 mm at introduction.

 

The spec sheet for the KAF-18500 is not available, so I have been using the KAF-10500 for my examples for the Bayer Filter. The M8 uses a 0.5mm IR Filter. that is quoted in the Spec sheet.

 

Basically, the Bayer filter of the KAF-10500 is 1 stop. I assume the KAF-18500 is about the same.

 

I would love to add the KAF-18500 spec sheet to my exciting collection of Kodak Sensor Data sheets that go all the way back to the KAF-1300. Interesting, The KAF-1300 did not make use of the Bayer pattern in it's sensor. It used a different Mosaic filter. The Bayer pattern was patented in 1976, by Kodak. Have to wonder why it was not used until the DCS200c.

[Advertisement]

Hallo,

Look here: Leica M

[thebarnman]

I used to play with colored filters when I was a little boy. And, it was really neat to see what happened when I put a blue filter on top of a red filter and look through it.

 

Wow, everything turned purple!

 

I also noticed things didn't look as bright while looking through the filters.

[giordano]

I think I've posted something like this before but I'll try again:

 

There seems to be an important difference between (a) the filters needed to get R, G and B images from a photographic subject (e.g. those in a Bayer array or an Autochrome, Dufaycolor or Polachrome film) and ( those needed to make colour separations from a tripack image (e.g. when printing a Kodachrome photo in National Geographic).

 

In ( the curves of the R,G and B filters ideally have zero overlap, so that each of the separations contains only the information from the corresponding layer of the emulsion (colour channel). In (, a good deal of overlap is not only acceptable but desirable because the aim is that the information in the R, G and B channels should broadly correspond to that sensed by the corresponding L, M and S cones on your retina.

 

This suggests to me that light loss from a Bayer CFA should be substantially less than from process filters.

 

What I haven't yet been able to find out is whether the "colour separation" filters used on photographic cameras were actually different from those used on process cameras. Does anyone know?