--being Assimilated, by...
The Borg.
A bit of the old, and a bit of the "newer".
Frankly, I am having brain burn trying to assimilate this stuff
myself. Too much new thinking for this old dog......ouch....
--
Bill Dewey
http://www.deweydrive.com
--Ed,
The lights used in many studio environments can add significant
levels of infra red radiation to a scene that though invisible to
the human eye is quite visible to the sensors used in video camera
equipment (which suprise suprise use CMOS and CCD based sensors
predominantly these days).
You need to recognize the mitigating factors to your use of these
filters and employ them in your shoots just as you would employ
neutral density, polarizing and UV blocking filters to modify
images to your liking.
1) The wavelengths produced by your light sources are important.
2) The reflectivity/absorption characteristics of objects in your
scene is important.
Recognize that certain light sources may have large frequency
components in the IR or UV. As you discovered in your shot about,
tungsten has components in the NIR it also has series in the UV and
little in the standard green/yellow frequencies that humans are
most sensitive to. See this spectrograph of the emission lines of
tungsten:
Note how most of the components are toward the ends of the visible
spectrum. Seeing the spectrogram, it should make sense that setting
proper white balance for your camera is critical to getting the
right appearance from your scene under tungsten light. Too much
balance toward the blue and your red components are lost, too much
toward the red and blue is lost...making the task of calculating an
optically accurate final pixel value for the illuminated material
very difficult. Tungsten lights correspond to about 2850 on the
Kelvin scale, about where you set your camera before shooting the
test so the major contribution of your IR noise is coming from the
reflected (absorbed and reemmited) components in the material of
the jacket. The color calculations for the pixels take the white
balance into account in order to determine the final pixel, if you
set the wrong color temperature it will calculate incorrect values
for the color of pixels. Additionally, the reflectivity or
absorption characteristics of certain materials may make the
situation worse, when you set white balance you are trying to match
your direct lighting from the sources (tungsten) however the
reflected spectra from the illuminated materials can be very
different from the emmission spectra of the light source making it
even more difficult for your sensor and CFA to be able to determine
an optically accurate final pixel color.
Without the hot mirror filter attached, try several shots with the
same scene under the same lighting and vary the white balance, you
should NOT see significant variation in how much purple/blue is
introduced into jacket in the final image(though the rest of the
scene will of course shift). Add several objects that under the
same lighting are also optically black but are composed of
different materials from the cloth in your suit. You should find
that some optically black materials are much better at absorbing
(and not reflecting) the IR from the light source and thus render
pixels in the image that are closer to the "black" of their
apperance to your eye than others. (Assuming the white balance is
set to match the light sources.) Many common day materials which
may appear the same color when observed with the eye appear to glow
when viewed in the IR field as they radiate those frequencies
emmitted from the surrounding light sources to varying degree while
absorbing all other optical frequencies that human eyes would be
sensitive to.
All that said, I think the D2h sensor and CFA combination is such
that under a combination of tungsten lighting and IR reflective
scene materials there is a high likelyhood of calculating an
incorrect (towards blue) pixel value which is particularly
noticeable over dark or IR reflective areas of the scene, this
shows up as pixels rendered with visible blue/purple components
(the yucky chroma noise in the shadows) Using the hot mirror filter
goes a long way apparently to re asserting the spectral balance and
allowing the calculated pixel values to fall under the correct
(optically accurate) colors. If you noticed in Bjorn Rorslett's
review he specifically mentioned that sensitivity to IR for the D2h
is lower than it was in the previous D series (his estimate by
almost 2 stops http://www.naturfotograf.com/index2.html ) which
corroborates the blue/purple noise of certain materials illuminated
under the tungsten sources, the CFA is cutting off half the
incoming illumination components (the lump of near IR red
components of the sodium emmission spectrum) leaving the lump of
blue components relatively alone to form the significant
contribution of the final calculated pixel. That's the best theory
I can muster on the issue so far I'd be interested in reading
alternative views.
If you are interested in the emmision spectra of other elements and
how they compare to Sodium look here:
http://home.achilles.net/~jtalbot/data/elements/
Regards,
--
--Ed,
The lights used in many studio environments can add significant
levels of infra red radiation to a scene that though invisible to
the human eye is quite visible to the sensors used in video camera
equipment (which suprise suprise use CMOS and CCD based sensors
predominantly these days).
You need to recognize the mitigating factors to your use of these
filters and employ them in your shoots just as you would employ
neutral density, polarizing and UV blocking filters to modify
images to your liking.
1) The wavelengths produced by your light sources are important.
2) The reflectivity/absorption characteristics of objects in your
scene is important.
Recognize that certain light sources may have large frequency
components in the IR or UV. As you discovered in your shot about,
tungsten has components in the NIR it also has series in the UV and
little in the standard green/yellow frequencies that humans are
most sensitive to. See this spectrograph of the emission lines of
tungsten:
Note how most of the components are toward the ends of the visible
spectrum. Seeing the spectrogram, it should make sense that setting
proper white balance for your camera is critical to getting the
right appearance from your scene under tungsten light. Too much
balance toward the blue and your red components are lost, too much
toward the red and blue is lost...making the task of calculating an
optically accurate final pixel value for the illuminated material
very difficult. Tungsten lights correspond to about 2850 on the
Kelvin scale, about where you set your camera before shooting the
test so the major contribution of your IR noise is coming from the
reflected (absorbed and reemmited) components in the material of
the jacket. The color calculations for the pixels take the white
balance into account in order to determine the final pixel, if you
set the wrong color temperature it will calculate incorrect values
for the color of pixels. Additionally, the reflectivity or
absorption characteristics of certain materials may make the
situation worse, when you set white balance you are trying to match
your direct lighting from the sources (tungsten) however the
reflected spectra from the illuminated materials can be very
different from the emmission spectra of the light source making it
even more difficult for your sensor and CFA to be able to determine
an optically accurate final pixel color.
Without the hot mirror filter attached, try several shots with the
same scene under the same lighting and vary the white balance, you
should NOT see significant variation in how much purple/blue is
introduced into jacket in the final image(though the rest of the
scene will of course shift). Add several objects that under the
same lighting are also optically black but are composed of
different materials from the cloth in your suit. You should find
that some optically black materials are much better at absorbing
(and not reflecting) the IR from the light source and thus render
pixels in the image that are closer to the "black" of their
apperance to your eye than others. (Assuming the white balance is
set to match the light sources.) Many common day materials which
may appear the same color when observed with the eye appear to glow
when viewed in the IR field as they radiate those frequencies
emmitted from the surrounding light sources to varying degree while
absorbing all other optical frequencies that human eyes would be
sensitive to.
All that said, I think the D2h sensor and CFA combination is such
that under a combination of tungsten lighting and IR reflective
scene materials there is a high likelyhood of calculating an
incorrect (towards blue) pixel value which is particularly
noticeable over dark or IR reflective areas of the scene, this
shows up as pixels rendered with visible blue/purple components
(the yucky chroma noise in the shadows) Using the hot mirror filter
goes a long way apparently to re asserting the spectral balance and
allowing the calculated pixel values to fall under the correct
(optically accurate) colors. If you noticed in Bjorn Rorslett's
review he specifically mentioned that sensitivity to IR for the D2h
is lower than it was in the previous D series (his estimate by
almost 2 stops http://www.naturfotograf.com/index2.html ) which
corroborates the blue/purple noise of certain materials illuminated
under the tungsten sources, the CFA is cutting off half the
incoming illumination components (the lump of near IR red
components of the sodium emmission spectrum) leaving the lump of
blue components relatively alone to form the significant
contribution of the final calculated pixel. That's the best theory
I can muster on the issue so far I'd be interested in reading
alternative views.
If you are interested in the emmision spectra of other elements and
how they compare to Sodium look here:
http://home.achilles.net/~jtalbot/data/elements/
Regards,
--
--
--
Ed, I appreciate your work here. And like Bill Dewey, I wonder how and when it would be most appropriate to use this. I also agree that this is probably something that Nikon development engineers could do well to investigate and implement. BUT, I do see a difference in sharpness between the two files. Where the pants in the filtered file are extremely sharp, the tie leaves a bit to be desired. In the unfiltered file, the tie looks great and the pants are full of red and quite soft. I can't help wondering if that couldn't be addressed with de-moire and a WB adjustment.I think this shows a clearer example of the effect a hot mirror has
on a D2H image. These pictures are of a pair of suit pants and
matching tie. The pants are BLACK and the tie BLACK and GRAY.
I'm a conservative dresser, no purple for me.
I won't tell you the one with the hot mirror, you'll have to scroll
down to see that.
Both images were shot under a single source tungsten light with a
white balance set to 2550K. The ISO is 1250 -The exposure times
were the same.
Someone wrote in a previous post that the difference wasn't really
noticeable - well, I guess it depends on your perception
Now, what I have to test is how the D1H and D100 do on this same test.
My web host is getting a little peeved about the amount of band
with I'm taking up, so I didn't post the full size links , in this
sample the side by side is a crop 100% of the full image which is
shown as a top and bottom comparison. If you really want a full
size, drop me an e-mail and I'll give you a link.
If you couldn't tell,
The image on the
LEFT and BOTTOM
are the ones with the Hot Mirror.
--
Ed Betz
http://www.edbetz.com
--
--
--
As expected, the filter has IR allowence. What was the exposition?Here is a photo if the light from my TV remote.
By the way, as expected, the same shot with the hot mirror filter
is just black.
Now, tell me again what this is going to show?
--As expected, the filter has IR allowence. What was the exposition?Here is a photo if the light from my TV remote.
By the way, as expected, the same shot with the hot mirror filter
is just black.
Now, tell me again what this is going to show?
Blue channel is affected. Can you e-mail me NEF? [email protected]
--
no text
http://forums.dpreview.com/forums/read.asp?forum=1021&message=6934689
