Kodak sensor

[bsmith]

The biggest mistake Leica made was partnering with Kodak ! As a 30 year Canon user , I had to switch to Nikon (D1) until Canon's contract with Kodak expired and they were able to put their own chips in their digital cameras. Even the Kodak/Canon $20,000 6mp camera had this same "blacks as purple " problem. The "hot mirror" filter helped some but never completely eliminated the purple cast. That was around 1998-1999 ! Why didn't Leica choose a different CCD maker ?

Leica HAD to be aware of this default. It's absolutley impossible to walk around with the M8 for even 5 minutes at night and not get banding. And at even ISO 160 outside , daylight balanced, shoot anyone, in a black tshirt or jacket and it will be purple. (and black cars or other non fabric elements in the same scene will be black)

If you carefully choose your floral or landscape scenes, yes, it's very good image quality.
But use the camera for 5 minutes in a "real world street scene" and it's a disaster.

[cbretteville]

Makes you wonder why the Kodak chip in the DMR records black fabrics as black....

- C

[carstenw]

Much thicker IR filter.

[bsmith]

Is it a Kodak chip ? I read somewhere it was made by Sony

[carstenw]

Kodak.

[jrc]

Originally Posted by bsmith

It's absolutley impossible to walk around with the M8 for even 5 minutes at night and not get banding. And at even ISO 160 outside , daylight balanced, shoot anyone, in a black tshirt or jacket and it will be purple. (and black cars or other non fabric elements in the same scene will be black)

This simply isn't true.

I've shot at night quite a bit the last few days, and have only a few cases of banding, and most of those, I forced. You can shoot for hours without gettting banding or severely limiting yourself in your shots.

Only a subset of black fabric turns color -- that subset that reflects more than average IR. A cotton t-shirt stays black (at least mine does), but my black raincoat has a magenta cast, because, I think, of the waterproof coating. But most of the time, shooting in dayight or night, blacks stay black. It's the unpredictability that hurts.

JC

[Vivek Iyer]

Since Leica is the company selling this camera, I would have thought that they would have checked the camera, its spectral response, etc instead of relying on goofed up Kodak data.

No point in blaming Kodak for the sensors. Except Canon and Sony no other players make their own sensors. Others rely on Kodak or Sony.

Sony sensors are used in (Minolta) Sony, Pentax and Nikon cameras. Olympus and Leica use Kodak's sensors. Cameras from differnt manufacturers using the same CCD sensor functions differently. So, for the buyers, it the camera manufacturer and not the chip manufacturer who has to bear the responsibility.

[bsmith]

Quote"

This simply isn't true.
I've shot at night quite a bit the last few days, and have only a few cases of banding, and most of those, I forced. You can shoot for hours without gettting banding or severely limiting yourself in your shots.

Only a subset of black fabric turns color -- that subset that reflects more than average IR. A cotton t-shirt stays black (at least mine does), but my black raincoat has a magenta cast, because, I think, of the waterproof coating. But most of the time, shooting in dayight or night, blacks stay black. It's the unpredictability that hurts."



Simply not true for you,...,. as far as my M8, most night shots with street lamps exhibit banding, look closely at the files. Maybe we shoot differently of your sensor is better than mine.

On the black fabrics,... almost all my cotton tshirts, all silk coats, fabric on office desk chairs, cotton pants, socks, black towels, black pillows, Rubber shoes, metal painted black, and black leather, wood painted black are all fairly close.

Read the values in PS, maybe your screen is off,

[Bob Ross]

Originally Posted by Vivek Iyer

Since Leica is the company selling this camera, I would have thought that they would have checked the camera, its spectral response, etc instead of relying on goofed up Kodak data.

No point in blaming Kodak for the sensors. Except Canon and Sony no other players make their own sensors. Others rely on Kodak or Sony.

Sony sensors are used in (Minolta) Sony, Pentax and Nikon cameras. Olympus and Leica use Kodak's sensors. Cameras from differnt manufacturers using the same CCD sensor functions differently. So, for the buyers, it the camera manufacturer and not the chip manufacturer who has to bear the responsibility.

What goofed up data did Kodak provide?
You left out Fuji, Foveon, Nikon, Dalsa and Panasonic as larger sensor mfrs.
Bob

[Vivek Iyer]

Bob, Nikon buys their sensors from Sony. They made one sensor for their D2H(s). No more.

Yes, I did leave out Dalsa (real big player) and Foveon (esoteric). Don't know where Fuji get their sensor from. May be they make it themselves. No clues about Panasonic and large sensors.

Look here: http://www.leica-camera-user.com/dig...you-think.html
for a screen shot of the spectral response provided by Kodak for the sensor used in M8. The whole pdf file is on Kodak's site.

If that spectral response is really valid, then the M8 should not suffer from so much IR realted problems. It is really not a software problem. Interesting that Leica folks did not test the camera before its launch.

[reddawn]

Originally Posted by Vivek Iyer

Interesting that Leica folks did not test the camera before its launch.

why should they even bother when they have paying customers to test for them, suggest fixes n workarounds, and even very patient and forgiving by putting up with the obvious flaws and buying hundreds of dollars of filters to temporarily migitate the problem?

[Bob Ross]

Originally Posted by Vivek Iyer

Bob, Nikon buys their sensors from Sony. They made one sensor for their D2H(s). No more.

Yes, I did leave out Dalsa (real big player) and Foveon (esoteric). Don't know where Fuji get their sensor from. May be they make it themselves. No clues about Panasonic and large sensors.

Look here: http://www.leica-camera-user.com/dig...you-think.html
for a screen shot of the spectral response provided by Kodak for the sensor used in M8. The whole pdf file is on Kodak's site.

If that spectral response is really valid, then the M8 should not suffer from so much IR realted problems. It is really not a software problem. Interesting that Leica folks did not test the camera before its launch.

Don't rule Nikon out of the sensor making field and I think they said it is possible. Fuji does make their own sensors. Panasonic made the CCD in one of the Canons (the one that had a CCD) and the current 4/srds sensor.
I did see the Kodak specs and the QE for red is interesting. To say that Leica didn't test the M8 before launch (or Nikon the D200) is a bit of an exaggeration at least. To say they (the testers) should have noticed things, that we are seeing, is another.
Bob

[robertwright]

interesting to compare the kodak data for the dmr (senor kaf 10100) and the M8 (sensor kaf 10500).

looking at the coverglass the dmr has transmission as follows
420nm 65% min
420-630nm 80% min
45% at 660nm
700 to 750nm (this is the IR) less than 5% max
700 to 1000nm (Ir goes up to 1000nm) less than 0.5% ave.
1000 to 1100 nm less than 4% max

substrate schott d-263

coverglass specs for M8

400nm >80%
420-530nm >90%
610-630nm 50% crossover point
710-750 <10%
750-1000nm <5%
1000-1100nm <10%

substrate Kyocera B-7

Seems that the schott coverglass is doing twice the work the Kyocera is doing.
the schott substrate is .76mm and the kyocera is .5mm

what follows is that that what leica is saying is correct, the Ir filter is thinner. And ergo, has worse Ir performance by a factor of two.

Now it is very hard or impossible to find specs on the canon chips, but in their own EOS system catalogue they describe the cmos sensor as having several layers, two low pass layers, infrared absorbing glass and most interesting, a dichroic mirror layer to reflect Ir. You can see it if you look at the chip, there is a red mirror like surface under reflection sometimes.

Seems that canon determined it was necessary to have a dichroic mirror layer in addition to the Ir absorbing glass.

More interesting is that the thickness of this dichroic filter layer has to be an odd multiple of 1/4 the wavelength to be blocked, so for example if the Ir to be blocked is 750nm, 1/4 of that is 187.5 times an odd multiple, say 3 layers, which equals 562 nm or about half a micron. even if it had to be 9 layers that is only 3 microns of coating.

So they could have coated the absorbing glass to further reduce Ir...or added a .25mm dichroic filter in front of the .5mm coverglass and still have only been at .75mm total, the same as the DMR.

something to ponder...

[ho_co]

Originally Posted by robertwright

... something to ponder...

Robert--
Very interesting figures. Thanks for posting!

--HC

[robertwright]

I think my speculation about coating the absorbing glass has its problems, from other threads and in reading about dichroic filtration it gets less effective the more off-axis you get. So while canon can use a dichroic filter on their sensors Leica cannot, the angle of incidence at the corners means that the dichroic filter would really not work. You would have good filtration at the center and progressively worse filtration at the corners-essentially impossible to correct. coding woudl make no difference.

There were obviously some pipers to pay for the rangefinder design, and having the lens so close to the CCD may have been stellar for film, but not so stellar for digital.

From the examples posted the hot filter on the front seems to fix the problem. Great! now we can move on to moire! just kidding, I hate that canon mushy pixel look.

[Riley]

Originally Posted by Vivek Iyer

Bob, Nikon buys their sensors from Sony. They made one sensor for their D2H(s). No more.

Yes, I did leave out Dalsa (real big player) and Foveon (esoteric). Don't know where Fuji get their sensor from. May be they make it themselves. No clues about Panasonic and large sensors.

Look here: http://www.leica-camera-user.com/dig...you-think.html
for a screen shot of the spectral response provided by Kodak for the sensor used in M8. The whole pdf file is on Kodak's site.

If that spectral response is really valid, then the M8 should not suffer from so much IR realted problems. It is really not a software problem. Interesting that Leica folks did not test the camera before its launch.

canon also use Sony sensors in their P&S
Fuji are special, the incredible F30 sensor is a new size 1/1.7" (perhaps to keep it for themselves) i believe they are inhouse
Panasonic make their own, labeled Matsushita
Sharp make sensors too

Riley

[robertwright]

BTW I was wrong, it gets more aborptive the more off axis you go, eventually getting into the visible spectrum....

I will have to reverse my thinking on this and say it would be a good thing to have this all inside the camera and not outside.

Originally Posted by robertwright

I think my speculation about coating the absorbing glass has its problems, from other threads and in reading about dichroic filtration it gets less effective the more off-axis you get. So while canon can use a dichroic filter on their sensors Leica cannot, the angle of incidence at the corners means that the dichroic filter would really not work. You would have good filtration at the center and progressively worse filtration at the corners-essentially impossible to correct. coding woudl make no difference.

There were obviously some pipers to pay for the rangefinder design, and having the lens so close to the CCD may have been stellar for film, but not so stellar for digital.

From the examples posted the hot filter on the front seems to fix the problem. Great! now we can move on to moire! just kidding, I hate that canon mushy pixel look.

[masjah]

Robert

If the absorption varies too much with position, surely you've still got exactly the same problem in reverse - that is, get enough absorption in the middle, and you may be cutting into the visible spectrum at the edges? Just a thought!

[pfogle]

Originally Posted by robertwright

BTW I was wrong, it gets more aborptive the more off axis you go, eventually getting into the visible spectrum....

I will have to reverse my thinking on this and say it would be a good thing to have this all inside the camera and not outside.

If the wavelength for maximum absorption is 4 times the path length, then it doesn't get either more or less absorbtive off-axis, but the wavelength for max. absorbtion (actually max. destructive interference, which is a bit different) will increase off axis, as the path length is longer for oblique rays. So, longer IR cut out in the corners, and shorter IR in the center. Visible should not be affected at all.

A couple of other things come to mind...
-the only other digital RF, the Epson, seems to also show some IR contamination, though not quite as much as the M8 - a different filter system perhaps? That camera uses the Sony chip as in the Nikon D100, I believe.
-I don't get this thing about the filter thickness - seems hard to believe they couldn't get an extra 0.25mm from somewhere in the design.
-The worst case might be a bit more vignetting with some wide-angle lenses. This is adequately dealt with in the R-D1 in the RAW converter, also in CS2, and in the M8 via lens coding and in-camera firmware. Surely they could just ramp up the correction a bit and put a better filter in?

A final point about the testing - when the Hubble was launched all those years ago, it had an out of focus secondary mirror because of a fault, not in the mirror, but in testing rig! Same might apply here - if they only tested the camera under lab conditions they might never have seen either the banding or the magenta cast in the test setup. It's an easy mistake for engineers to make - the lab is not the real world.

-edit- I just realised the last comment is probably nonsense, as the camera had been beta tested in the field - maybe they just weren't listening...

[Vivek Iyer]

Originally Posted by pfogle

If the wavelength for maximum absorption is 4 times the path length, then it doesn't get either more or less absorbtive off-axis, but the wavelength for max. absorbtion (actually max. destructive interference, which is a bit different) will increase off axis, as the path length is longer for oblique rays. So, longer IR cut out in the corners, and shorter IR in the center. Visible should not be affected at all.

Quite right.

A couple of other things come to mind...
-the only other digital RF, the Epson, seems to also show some IR contamination, though not quite as much as the M8 - a different filter system perhaps? That camera uses the Sony chip as in the Nikon D100, I believe.

Different filter. But really not as good as the one in M8. The filter in M8 is AR coated on both sides. In this sense, it is excellent.

-I don't get this thing about the filter thickness - seems hard to believe they couldn't get an extra 0.25mm from somewhere in the design.

The thickness is a non issue. They (Kodak) could have used a stronger filter.

-The worst case might be a bit more vignetting with some wide-angle lenses. This is adequately dealt with in the R-D1 in the RAW converter, also in CS2, and in the M8 via lens coding and in-camera firmware. Surely they could just ramp up the correction a bit and put a better filter in?

The major improvement in the Kodak sensor over the Sony one is the use of micro lenses to correct for light fall off. Nothing to do with filters.

A final point about the testing - when the Hubble was launched all those years ago, it had an out of focus secondary mirror because of a fault, not in the mirror, but in testing rig!

Quite right. They didn't test or listen well. The Hubble fiasco came about mainly because of feet dragging. NASA/contractors should have converted to metric, at least after this major/costly blunder.