Using D300 to photograph a visual Hydrogen spectrum

[mike_cowan1]

For a demo/lecture on light and colour, I wanted to photograph a hydrogen spectrum using a narrow hydrogen lamp,

a slit, and a grating. The resulting spectrum was very clear by eye - the central pink line of the lamp itself,

one deep red line, one green line and one deep violet line. When photographed with a D300, the image was the

same central pink line, but the red line was nearly orange, the green line was a light clear blue, and the violet

line was blue with a violet tinge. No amount of white-balance, exposure or ISO shifting could turn the green

line (admittedly a blue-ish green, but still clearly green) into any colour but nice bright blue. A design

problem in the colour sensors? A problem with the extremely narrow wavelength range of the spectral lines

themselves? A flaw in the camera's colour processing? I'd be grateful for any suggestions you have.<div></div>

[ellis_vener_photography]

This really seems like a question for Bjorn Rorslett !

 

Are you shooting NEFs or JPEGs?

[rffffffff]

One dumb suggestion: Many times the color of a light wobbles a bit over time... I seriously doubt this is the case here, but I would try a long shutter speed to rule out that particular problem.

[geoffs1]

I suspect this is an effect related to Metamerism: http://en.wikipedia.org/wiki/Metamerism_(color)

 

The responses of the sensor elements and their fairly broad filters may not generate exactly the same response as

the the cones in your eye. Monochromatic colors are the most problematic because they show very small

mis-matches between the red, green, and blue sensor response curves.

 

FWIW, 486nm is generally considered "blue-green", and it's often photographed more blue than green, for example:

http://hyperphysics.phy-astr.gsu.edu/hbase/hyde.html#c4 and

http://hyperphysics.phy-astr.gsu.edu/hbase/Tables/hydspec.html and

http://www.chemguide.co.uk/atoms/properties/hspectrum.html

 

You might try looking at the RGB values for each line to see if you can create a WB or RGB set of curves to

reproduce what you see. Raw would probably help.

[mike_cowan1]

Ellis, I'm shooting both NEF's and JPEG's, and they're identical in colour.

 

Robert, this was a hydrogen discharge lamp, with wavelengths fixed by the physics of hydrogen. In any case, I tried it maybe 20 times under different conditions, and it was always the same.

[mike_cowan1]

Geoff, thanks very much for the references. They all show the 486nm line as much bluer than it appears to the eye, so apparently I'm not alone in this. In Photoshop I had to select the three lines individually and use 'Hue and Saturation' different amounts with each one to make an image that looked roughly like what I saw. OK for hydrogen, but way too much trouble for a complicated spectrum! Thanks again.

[iliafarniev]

I think it is normal reaction of CCD on clean and narrow spectral light of given wave length. I would try to shoot same with another digital camera, pref. not a Nikon, to see if it make a difference. How this spectrum is actually obtained? Are you photographing thorough a spectroscope or project thorough the glas prism?

[ellis_vener_photography]

HAve you tried using a gray background instead of a black background/ I t may be that the BAyer color interpolation is just not picking up enough color data to do a more accurate interpolation. But really I'm over my head on this one. Bjorn does a lot of work with infrared and UV light however so he may be of assistance or interested in the challenge

[Richard Williams]

Bill Claff posts on Nikonians and has a very good handle on how Nikon dSLRs process colour:

 

http://home.comcast.net/~nikond70/

 

He might well be interested in your issue - drop him a line.

[bill koenig]

Doesn't the D300 us a CMOS senser? The D200 does use a CCD

[geoffs1]

Mike Cowan, "Geoff, thanks very much for the references. They all show the 486nm line as much bluer than it appears to the eye, so apparently I'm not alone in this."

 

I think it's a quite general characteristic of RGB-based color imaging processes. It would be interesting to see how different color films record the colors. With negative films you would have the added complication of color "correction" during printing that reversal film wouldn't have.

 

An alternative processing for a more complex spectrum might be to convert the image into b/w and use it as a mask/overlay for a "color spectrum" gradient layer that contains the colors you want to see along the spectral axis. For any fixed spectrograph/camera setup the gradient would be the same.