[Sticky] Star Color Calibration 1.58 - an example using Frank Schmitz's M101 composite with complete explanation of the tool

Hi Frank, @foschmitz

Thank you for sharing your data.

First my compliments on great M101 data 😉 

I am too blame here, since I haven't yet provided instructions on this tool 🙄 but having said that, this is a very nice example to correct this now. So I will start here with giving a thorough explanation.

The Black Body calibration model is based on real physics for the star colors of Main Sequence stars:

Most of the stars in our images are Main Sequence stars in our own galaxy, the Milky Way. These stars are in the first phase of their live, fusing hydrogen nuclei (protons) to helium nuclei in the core of the star. This phase is usually the longest phase in a star's life/evolution from a cloud of gas to a white dwarf (which is a dying star in fact) or super novae for instance. During this main sequence phase, the stars are in hydrostatic & thermal equilibrium. This equilibrium is the reason for thes stars to emit light almost identically as a Black Body on broadband wevelenghts.

Furthermore, it's important to realize the following 2 feautures of Main Sequence stars:

1. by far most stars are Reddish (95-99%). These are low-mass stars that have live relatively very long. Only a fraction (1-5%) of the stars will be blueish, these stars are high-mass stars that live relatively very short.
2. A red star has a low luminosity, a blue star has a high lumninosity. A blue star's total luminosity can be more  than 1.000.000x !!! as bright as a red star's total luminosity.

These 2 aspects have an important consequence !

If you look at a (main sequence) star population from a very great distance over which you can hardly see the individual stars anymore (like looking at a neighbour galaxy or M101 in this example). Then, although there are much more red stars, the blue stars will easily outshine the red stars in large parts of the population.

From Wikipedia:

https://en.wikipedia.org/wiki/Color%E2%80%93color_diagram

Although stars are not perfect blackbodies, to first order the spectra of light emitted by stars conforms closely to a black-body radiation curve, also referred to sometimes as a thermal radiation curve. The overall shape of a black-body curve is uniquely determined by its temperature, and the wavelength of peak intensity is inversely proportional to temperature, a relation known as Wien's Displacement Law. Thus, observation of a stellar spectrum allows determination of its effective temperature. Obtaining complete spectra for stars through spectrometry is much more involved than simple photometry in a few bands. Thus by comparing the magnitude of the star in multiple different color indices, the effective temperature of the star can still be determined, as magnitude differences between each color will be unique for that temperature. As such, color-color diagrams can be used as a means of representing the stellar population, much like a Hertzsprung–Russell diagram, and stars of different spectral classes will inhabit different parts of the diagram. This feature leads to applications within various wavelength bands.

The graphs on the left side of the tool are called color-color diagrams and are quite common in the scientific literature:

https://en.wikipedia.org/wiki/Color%E2%80%93color_diagram

In astronomy, color–color diagrams are a means of comparing the apparent magnitudes of stars at different wavelengths. Astronomers typically observe at narrow bands around certain wavelengths, and objects observed will have different brightnesses in each band. The difference in brightness between two bands is referred to as color. On color–color diagrams, the color defined by two wavelength bands is plotted on the horizontal axis, and then the color defined by another brightness difference (though usually there is one band involved in determining both colors) will be plotted on the vertical axis.

Now the key to using the Star Color Calibration tool is to look at the 2 graphs on the left side and the interpretation on where the stars are in the graphs and how these graphs work and knowing what astronomical color is.

Astronomical color:

The astronomical color B-R is the magnitude difference between the Blue and Red luminosity of the star. So what does this mean:

•  a value of B-R = 0 means that Blue and Red are equally bright for that particular star.
• If the value is larger then 0 (zero) then, now this is important ! since magnitudes are used, then the Red luminosity is greater than the Blue luminosity of that star. ( vice versa, a negative value means that Blue is more luminous than Red.)
• astronomical colors are always presented in such a way that lower values mean the star is more blue, and higher values mean, the star is more red. So that is a good rule of thumb to remember.

Let's look at the B-R versus G-R graph:

Look at the range of the G-R and B-R values for the plotted star population.

• The G-R values are all positive, meaning all the stars are more red than green.
• The B-R values are all positive, meaning all the stars are more red than blue.

So seeing a calibrated image with only red stars is confirmed by these graphs. Obviously it doens't fit our expectation of how the data should look calibrated. We expect some blue in there, right?

Blue-Red, slider:

To shift the star population towards blue, you can adjust the Blue-Red, slider. Bear in mind, if the star population selected is sufficiently large, we still expect more red than blue stars.

A value of -0.25 (which corresponds to the axis of the graphs) seems appropriate for your data. Basically, by adjust this slider, you adjust the B-R color of the average star in the population. Setting it lower means, the average star becomes more blue, setting it higher, means the average star becomes more red.

Note the range now of the B-R and G-R values.

• B-R range -0.40 to +0.60
• G-R range -0.23 to +0.40
• the bulk of stars have B-R colors of -0.2 to + 0.6. Most stars are still red, but we have some blue stars.

This slider can also be called the white star-calibration slider, since this enables you to adjust where a white star in the population should be located in the graphs.

Looks immediately much better after only 1 adjustment, right?

Further finetuning is in my next post...

Mabula

Awesome description Mabula, that helped a lot to understand! Very powerful tool you have created there, thanks a lot! Looking forward to the video tutorial.

Best Frank

Further finetuning can be done with the 2x slope & 2x constant sliders for the 2 color-color diagrams.

The slope and constant are 2 parameters of a linear function: y = ax + b

a = slope

b = constant

x = astronomical color on x-axis

y = astronomical color on y-axis

So the slope and constant values determine the Black Body line in the color-color diagrams.

The actual values depend on the filters that are used in acquiring the data. If you always use the same filters and camera, then you expect these slopes and constant to be the same for different objects.

The physics behind this is given by F.J. Ballesteros 2012, New insights into black bodies, https://arxiv.org/pdf/1201.1809.pdf .

The B-R versus G-R graph after the initial Blue-Red adjustment, already gives a hint. we can see that slopes are probably smaller.

B-G versus G-R slope = 0.75

B-R versus G-R slope = 1.50

That leaves us with the 2 constants. Lowering the constants will lower the Black Body model/line in the color-color diagrams and this will have the effect that the colors become more magenta. Vice versa, increasing the constants, will make the colors more green. The expected value depends, (as mentioned before, just like the slopes) on the actual filters that were used in acquiring the data. But reasonable values for most filters are between -0.50  to +0.25.

The initial/default values that were used are -0.20 for both constants. Setting these both at 0,0 produces a very fine result, both visually and in the color-color diagrams:

And to show that the found Black Body model for star color calibration does work on other stars in the field of view, giving a nearly identical result, I placed the area select boxes differently. Off course minor adjustments in the B-R slider can be expected for white star calibration, but not for the other sliders since they should be dependent on the used filters.

Finally, the Star Rejection Kappa slider is a slider that determines which stars are used to calculate the model. The higher the kappa, the more stars are used. Setting this at 2 kappa, the default value, will have the effect that about 95 % of the stars is used and outliers (which could be non-Main Sequence stars like white dwarfs, Wolf-Rayet, unstable, variable stars, stars transitioning from main sequence to a red giant, etc..) are not. Leaving it at 2, should prove fine in almost all cases. But setting it a higher value will have little influence, since the possible outliers will contribute little to the model calculation. As an example, I raised it to 3 kappa, which normally corresponds to using more than 99% of the stars.

To end this explanation on Star Color Calibration using a Black Body model, I show a crop of Frank's M101, the Pinwheel Galaxy. I think it's awesome with good colors, and Frank (@foschmitz) acquired great data, that always helps 😊 The amount of red color in the active star forming regions in M101 is stunning.

Kind regards,

Mabula

Sticky!?

Posted by: gdwats@comcast.net

Sticky!?

Yes, sticky 😉

Thank you so much for posting this. I am about to ask the same question. 

hello,

I am trying with this and it ain't easy. If I only slightly shift the value of -0.35

by 0.01 I get lots of green, see the attached screenshots. How can I make it look like Rogelio's (concerning colors)?

It is part of the mosaic Mabula showed on his talk ... I put a link to the fit data processed with APP 1.058

fit_by_APP

Stefan

Hi Stefan,

The slopes can probably be a bit steeper/higher, but they are almost okay.

Lowering the constants will have the effect of reducing green. You have now set it at -0.05 for both graphs. How does it look if you lower it to -0.20 for instance?

Make sure that most stars are red 😉

Mabula

Posted by: Mabula Haverkamp - Admin

Hi Stefan,

The slopes can probably be a bit steeper/higher, but they are almost okay.

Lowering the constants will have the effect of reducing green. You have now set it at -0.05 for both graphs. How does it look if you lower it to -0.20 for instance?

Make sure that most stars are red 😉

Mabula

Hi Mabula,

so it is that easy as understood, you shift until you have it right in the intervals suggested by you, then adjust the two constants that it visually fits to the black body data, that is all?

Do you have an idea why there is such a strong shift in the color by only changing the shift by 0.01, as seen by the two screenshots?

Yes, red, haha, I h&a%t§e green in stars...

Hey, any chance the mosaic tutorial will come soon? 😍  😍  😍  🥂 

Stefan

Posted by: elgol

Posted by: Mabula Haverkamp - Admin

Hi Stefan,

The slopes can probably be a bit steeper/higher, but they are almost okay.

Lowering the constants will have the effect of reducing green. You have now set it at -0.05 for both graphs. How does it look if you lower it to -0.20 for instance?

Make sure that most stars are red 😉

Mabula

Hi Mabula,

so it is that easy as understood, you shift until you have it right in the intervals suggested by you, then adjust the two constants that it visually fits to the black body data, that is all?

Do you have an idea why there is such a strong shift in the color by only changing the shift by 0.01, as seen by the two screenshots?

Yes, red, haha, I h&a%t§e green in stars...

Hey, any chance the mosaic tutorial will come soon? 😍  😍  😍  🥂 

Stefan

Hi Stefan,

Thes big shift with small paramters change, yes, it's something in the algebra. I am working in it. It has to do with initial parameters for the calculation.

Mosaic tutorial using your data is coming 😉

Mabula

@elgol,

Fixed the algebra problem, should work much better in 1.059. Starting parameters before were generally the same for all images. Now initial parameters are determined per image before calculation, which should be much better.

Mabula

Posted by: Mabula Haverkamp - Admin

@elgol,

Fixed the algebra problem, should work much better in 1.059. Starting parameters before were generally the same for all images. Now initial parameters are determined per image before calculation, which should be much better.

Mabula

Mabula, maybe your nickname is Scotty the miracle chief engineer of NCC 1701 ... 👍 🖐️ (hey, some emojiis are missing).

Should I wait for 1.059? Guess so is better ...  🤪 

Posted by: elgol

Posted by: Mabula Haverkamp - Admin

@elgol,

Fixed the algebra problem, should work much better in 1.059. Starting parameters before were generally the same for all images. Now initial parameters are determined per image before calculation, which should be much better.

Mabula

Mabula, maybe your nickname is Scotty the miracle chief engineer of NCC 1701 ... 👍 🖐️ (hey, some emojiis are missing).

Should I wait for 1.059? Guess so is better ...  🤪 

@elgol, did you try in 1.059 ?

Cheers,

Mabula

Hi Mabula,

not yet, but soon. Was hoping to receive the tutorial also, so I'd take one or two days off reprocessing all my Namibia data then  😆  😆  😎 

Grüsse,

Stefan