What colour are hydrogen emissions?

[MarkS]

Everyone "knows" the colour of hydrogen emissions don't they?  You just add your H-alpha data into your red channel and it's job done!

Yes it's true that H-alpha is red but emissions from hydrogen gas contain H-beta and H-gamma as well.  The colours of those emissions can be found on my annotation of a standard colour diagram below:



The colours of single wavelength spectrum lines are conveniently marked around the boundary of this chart.

So, depending on the mix of H-alpha, H-beta, H-gamma emerging from the cloud of dust you could potentially obtain any colour within the triangle marked, including bright pink!   If you add the wavelength of oxygen into the mix then you get an even greater possible range of interesting colours emerging from clouds of dust in our galaxy and beyond.

It's fascinating stuff and it's why I love full colour imaging.

Roger Clark explains a few more effects here:
http://www.clarkvision.com/articles/color.of.nebulae.and.interstellar.dust/

Mark

[Mac]

As you said add oiii which is 560nm up in the green and its a large triangle, giving you your RGB,
and not the HUBBLE PALLET, which is generally map whatever colour you like to whatever colour you want 

(I cant stand the Hubble pallet but thats another thread.)

True you can get any colour within that triangle, but the filters we use have a specific bandwidth,
so you will never get any blue light within the Ha filter,

So another question to raise is,
what bandwidth of filter should we be using?

if we use 3nm then that would effectively have very small rings around your RGB points, effectively missing the colour content as it falls away.
looking at the color graph, technically we would need filters with a bandwidth of 100nm each to allow for an overlap, but then you will get you Sodium emissions included from the light pollution at about 580, there must be a trade off at some point of bandwidth against true colours.
Ha will always be Red, Hb Blue and Oiii Green. but how much of the detail that falls outside of the bandwidth is being lost?

Im not sure if anyone has a complete collection of 3nm 7nm & 14nm filters for Ha, Hb & Oiii, but it would be interesting to do a comparison of say M42 in all of the bandwidths.

Worth a thought.

Mac.

[MarkS]

I wasn't talking about using filters at all but simply doing full colour (RGB) imaging.  The hydrogen emission colours are simply in the context of a wider full colour image.

Mark

[RobertM]

Thanks for bringing this and the other topic up Mark.  They are both very interesting and intertwined subjects.

You're right a lot of people 'just add Ha to red' but that may be the right thing to do with some nebulas but not (most) others.  The 'usual' thing is to add 20-25% to blue which is simple for Astro (sharp cutoff) RGB filters but it's a more difficult problem for (the more realistic) rgb filter arrayed sensors.  As for rgb array'd cameras, they are specifically designed to capture broadband colour information in the same way as the human eye.  The specific chart mentioned in the first post covers the response of the human eye but I thought that colour cameras would only record a subset of that due to their internal translation to Adobe RGB or the more restrictive sRGB colour spaces.  Interestingly all the Lyman Hydrogen emission lines are outside these colour spaces so I imagine that boosting the deep red transmission by removing the HP filter will bugger up the translation matrix (mentioned in the other post/CN).

It does pose an interesting problem -  all stars and dust clouds (illuminated by starlight) are broadband targets, should we then be using the same cameras for specific emission wavelengths (narrowband) or merging from monochromatic sources ( I think Mac was alluding to this below ?).  I Mac also produced an interesting post about merging NB data which might be relevant here.

Another question is how the sharp cutoff of individual colour filters in a mono camera (where there is virtually no crossover in spectral response between filters) compares to CFA sensored camera - one for another rainy day me thinks !

Sorry no conclusions, only adding (confusion?) to the debate.

Robert

[MarkS]

Ah yes - I think I can see what Mac was meaning - when merging data from a narrowband source we could add it into the image by giving it the correct hue.

I am going to investigate with a colour chart how much a camera modification mucks up the colour translation matrix.  I suspect it isn't much, at least if you are using a fairly sharp cut off beyond H-alpha as I do.  After all the Nikon D810a camera performs this trick quite well - it is H-alpha enabled but it gives reasonable colours for daytime terrestrial photography.  DXOMark has not yet produced a matrix for the D810a and it is not yet supported by DCRaw so I haven't had a chance to look into how it might differ from that of the D810. Remember that the colour translation matrix is also a compromise.  To reproduce colours more precisely it is necessary to photograph a colour chart under the lighting conditions you are using and an additional LUT (colour lookup table) is generated from this.  I think this might be a bit over the top for astro-imaging.  After all, we are not attempting to colour match pieces of artwork or fashion shoots.

It is quite true that the emission lines (as well as most spectral wavelengths from the rainbow) are actually outside most colour spaces - see the excellent diagram here for example.
https://en.wikipedia.org/wiki/ProPhoto_RGB_color_space

When I was playing around with the equations, converting from one space to another, I frequently encountered negative intensity values (for colours outside the gamut, including H-alpha).  Such values are usually clipped in order to be displayed.  It does mean that emission lines have to be mapped to a colour that is not identical to what the human eye sees.

Mark

[Mac]

To reproduce colours more precisely it is necessary to photograph a colour chart under the lighting conditions you are using and an additional LUT (colour lookup table) is generated from this.  I think this might be a bit over the top for astro-imaging.  After all, we are not attempting to colour match pieces of artwork or fashion shoots.

True we are not trying to colour match artwork, but this is the problem, we are taking an image where we do not know what the original photo looks like or what the final photo should look like,
yes we know that we have R G B colour data and we normally produce an image that is pleasing to us.

If you did start with the colour chart data and did manage to produce the correct numbers for a perfect replica of the test chart that still leaves one problem.

We still do not know what the original (space) image should look like.
How much absorption of the individual RGB images has taken place ect, take M42 for example,
most of us have photographed it and there are many fantastic images showing the dust, colours ect, and there are as many different hues of the Ha as there are images,
the only colour chart that you have to go on are the central stars which we know the temperature of and thus we know the colour of.

again the problem here is we can photograph those stars so that they are not saturated and then produce the numbers so that when stacked ect central stars are the correct colour temperature but how does this then expand when combining the longer exposures for all the fine detail.

Q) When you photograph the colour chart under a normal exposure and check the numbers you will get one set of values to allow you to produce the correct image, how does this fare when taking 60 low light R G B images ect as we do in normal astronomy, then stacking and then producing the numbers, I personally would have though the numbers would be identical as the original colour data technically has not changed, but i have a feeling that they wont be.

Might be worth an experiment.

Mac

[MarkS]

I would argue that the colours of M42 actually come out reasonably consistently across various (unmodified) DSLR cameras when processed with a daylight white balance.  That's despite the face that the single wavelength emissions are all outside the sRGB colour gamut.  The "red" in M42 typically comes out as a kind of pink indicating more than one wavelength of hydrogen emission. 

Mark