New star generator for other galaxies... (work in progress)

[Cham]

Easy !

What is this about ?

guess.jpg

Okay, this is work in progress...

[Hungry4info]

M31's central binary black hole?

[Cham]

M31's central binary black hole?

It was already done, long time ago.

Guess again !

Well, here's a clue (easy !), since you're talking about this object...

pict1.jpg

pict2.jpg

This is just a prototype for now...

EDIT : I must admit that I'm pretty excited by this new toy ! Space feels soooo vast now, geez this is impressive !

[Hungry4info]

The M31 galactic core?
Yeah I know that's little different from the first guess. All I know is you're in M31, and in its inner parts, and there's a bunch of stars.

[Cham]

The M31 galactic core?
Yeah I know that's little different from the first guess. All I know is you're in M31, and in its inner parts, and there's a bunch of stars.

You're close !

EDIT : Some progress (hmmm, need more work on this, but it's going well...)

pict0.jpg

pict4.jpg

[Hungry4info]

You're populating the entire galaxy? How many stars?

[Cham]

Yes !

I'm building a full star generator. I intend to put around 50k stars in the Andromeda galaxy (and maybe some others too). I want the population to be reasonably realistic (the stars should give a proper HR diagram, in my HR LUA tool). For the moment, just main sequence stars.

On the pictures above, the magnitude is grossly exagerated, so I can see the whole distribution while tuning the parameters.

The disk distribution should also be properly oriented with the galaxy model in Celestia.

[Hungry4info]

Very nice, very nice. I've considered doing something like this before, and adding randomly generated planetary systems for them.

[Cham]

Very nice, very nice. I've considered doing something like this before, and adding randomly generated planetary systems for them.

Well, my star generator doesn't add planets. Just plenty of stars on the main sequence.

More later...

EDIT : Here's an HR diagram (from the LUA Tools) showing the distribution of a sample of 2000 stars from my new generator. Maybe the curve needs some adjustements.

Stars from other generators (the old generator from Rassilon, for example) doesn't give realistic distributions, on the HR plot.

HR.jpg

[Cham]

Now I have a question for everyone : how should I name all these fictious stars in the Andromeda galaxy (M31) ?

Currently, they are named as "And" followed by a random number.

I could add a set of names randomly selected (with numbers), as this :

"And 124569"
"M31-34672"
"Toto 35678"
...


Suggestions ?

Once done properly, the STC files (various numbers of stars) will be published...

[Cham]

I got the orientation right, by trial and error (I still need a systematic way to do this, to ease the creation process to other galaxies in our local group).

The stars luminosity is grossly exagerated on the pictures below, so we could see the whole distribution. Currently, there are 10 000 stars in there. I could very easily change this to 50 000 stars globally (I want more in the central core). 10k stars gives an STC file of about 1.4 MB.

I have to tweak a bit the HR curve for more "realism", to change a bit the statistical distribution on the galactic thickness (especially for the central core), and to choose better star names (what are your suggestions ?).

side.jpg

3dview.jpg

People would have fun to add all sorts of Science-Fiction addons in-there !

[Cham]

I made some progress on the shape of the distribution. Again, the luminosity of stars shown below are greatly exagerated, so we can see the whole distribution. The AbsMag will be decreased to normal values when I'll be satisfied with the distribution) :

Front side (there are more stars on the bulge) :

M31a.jpg

Profile :

M31b.jpg

The distribution with the standard M31 galaxy :

M31c.jpg

There are 20 000 stars here. The distribution follow a classic main sequence curve on the HR diagram.

Before release, I need suggestions for the names. I don't think that it's a good idea to name all the stars "And ######" (where ###### is a random number).

[selden]

Personally I don't see any problem with using numbers as the suffix. I'd suggest using a prefix not related to the name of the constellation, though. That makes them too much like catalog designations for real stars.

[t00fri]

Cham,

I think your algorithm of generating (galaxy) stars is basically not correct, apart from their HR color distribution. There are several better attempts already in the net. In the first place, one has to take the basic inhomogeneity in star density into account that corresponds to the galaxy arms. Here is a close-up of the central quadrant of your M31 galaxy image.

cham.jpg

Your star distribution is locally almost homogeneous, while the galaxy "arms" are still consisting of my nebula sprites blended in.

In contrast, here is e.g. an interesting site introducing a generator for (spiral) galaxies in terms of "stars only":

http://dmytry.com/galaxy/index.html
(Dmytry Lavrov's, a mathematically oriented software dev)

The galaxy generator may be downloaded freely. Unfortunately, it's not open source. A simple example thereof illustrates pretty clearly what I mean:

galgen.jpg

It's straightforward to derive a 2D probability distribution with an arm structure, such that the stars distributed according to it follows the right pattern. It can be easily coded with Von Neuman's well-known acceptance-rejection method that I have also heavily used in my galaxy (and globular) code for Celestia. A nice summary of that method is here, for example,

http://pdg.lbl.gov/2011/reviews/rpp2011 ... niques.pdf

Your star distribution might be useful, however, for modelling elliptical and irregular galaxies as well as nearby spheroidal dwarfs that all are lacking any arms. Also never forget to consider distributing your stars according to a Perlin noise function...

The best approach I can see is a blended one, depending on the actual screen resolution or -- equivalently-- the observer distance from the galaxy. As long as the observer is far away, a sprite-based rendering is fast and perfectly sufficient. Upon a closer approach, when the resolution improves a lot, a blended transition to individual stars should be done. A big advantage is that at close distance from the galaxy, one only needs to render a tiny portion thereof in terms of stars, since the rest is outside the field of view! With clever culling, this fact might make a real-time rendering possible...

Such a two-component rendering has to be set up in the core code, since it is clearly not feasible via add-ons (at least in Celestia).

Fridger

[Cham]

Fridger,

yes I agree with you, and I was also thinking about the arms structures. I may do something about it next.

However, even with 50 000 stars, once distributed on the whole Andromeda galaxy (which is even larger than the Milky Way), the density is reaaaally low ! With proper AbsMagn, we don't see much the stars (except the highly luminous O stars) while moving through M31 ! It's even a bit deceptive, actually !

Don't forget that I extremely exagerated the luminosity of these stars to see the whole distribution in Celestia, just to orient the distribution correctly relative to Celestia's model. Their luminosity (AbsMagn) will be adjusted to normal values before releasing the STC files.

From 50 000 objects on a total of a billion of stars in reality, it's just a very low sampling (a proportion of about one star for one million) ! Any structure (with or without the arms) doesn't matter at all in this case.

But I do agree that making arms would be interesting. It's a lot of work to make them fit with the model in Celestia, though !

EDIT : There's an interesting unexpected effect, while moving through the distribution : we really feel the vastness of space in Celestia with these stars. Without them, I always add the impression that the Andromeda galaxy was just a small model in Celestia (even if it's to scale !). I never had the feel of a very large space. In other words, the stars give a sense of scale which was absent before. This fact alone makes the addon really worth it !


About the names, what do you recommend, Fridger ? I'm thinking about a random selection of words, followed by a number. But which words ? Currently, it's "And ######". There's a constraint : the user should be able to select a star in M31 using the name in Celestia, while knowing by its name that it's a fictious one.

[t00fri]

Cham,

Fridger,

yes I agree with you, and I was also thinking about the arms structures. I may do something about it next.

However, even with 50 000 stars, once distributed on the whole Andromeda galaxy (which is even larger than the Milky Way), the density is reaaaally low ! With proper AbsMagn, we don't see much the stars (except the highly luminous O stars) while moving through M31 ! It's even a bit deceptive, actually !

Since I have also experimented with placing stars in remote galaxies and the MilkyWay
http://forum.celestialmatters.org/viewtopic.php?t=210
I am aware about the low luminosity problem...

From 50 000 objects on a total of a billion of stars in reality, it's just a very low sampling (a proportion of about one star for one million) !

Therefore, one effectively will be forced into a 2-component approach, i.e. to render only a tiny portion of the whole galaxy in terms of say 50k stars when the observer is getting really close.

But I do agree that making arms would be interesting. It's a lot of work to make them fit with the model in Celestia, though !

With the blending technique in OGL, it's not really hard. One option would be to map the 2D sprite brightness to starnumber. Hence the bright portions of each sprite would be associated with many stars on close distance. Dark regions of the sprites correspond to a depletion of stars.

About the names, what do you recommend, Fridger ? I'm thinking about a random selection of words, followed by a number. But which words ? Currently, it's "And ######".

If you plan to read these stars in as real Celestia stars (via a STC file), then I am the wrong person to ask. There is also the option of generating the stars internally, exactly as I did with my globular stars. Then one doesn't need any names, but on the downside, one cannot land on them etc.

Fridger

[Cham]

If you plan to read these stars in as real Celestia stars (via a STC file), then I am the wrong person to ask. There is also the option of generating the stars internally, exactly as I did with my globular stars. Then one doesn't need any names, but on the downside, one cannot land on them etc.

Yes, my project is about "real" STC stars (I'm not against your proposed "fake" stars idea, like what we currently have for globulars. This is another topic).

One advantage of my STC project is that the SF fans could put all their stuff in M31, instead of "polluting" our Milky Way ... There's plenty of space (and now stars) available there !
Like what I said above, another advantage is the new feel of the vastness of space in M31. "Fake" stars (as in the globulars) aren't as effective, to give this sense of very high volume of space, because you can't select one of them and get closer (so the sense of scale).

[Fenerit]

Therefore, one effectively will be forced into a 2-component approach, i.e. to render only a tiny portion of the whole galaxy in terms of say 50k stars when the observer is getting really close.

FYI, such approach get the sense of vastness from outside the galaxy; whist once entered in it, "real" stars are good as well. As for star' names, I suggest to posit the prefix "Fict" (fictional) before And.

[Cham]

As for star' names, I suggest to posit the prefix "Fict" (fictional) before And.

Good idea. How about several names, just to add some variations ? (it may be dull to travel through the Andromeda galaxy and see all these stars called "Fict 2345", "Fict 12677", "Fict 68932", and so on...

I was thinking about something like "Fict Zeta 34521", "Fict Beta 678903", "Fict Chi 123212", etc...

Or maybe add the spectral type in the name : "Fict O 23456", "Fict M 231456", "Fict G 437896", etc...

EDIT : Hmm, I just had this idea : "And G234567", "And M78321", "And O5643", "And K03456". This way, we know the star is in Andromeda, and we know it's spectral class before selecting it. Maybe adding an "f" at the end to tell it's a fictious star (or is it too cryptic ?) : "And G21345f"

[t00fri]

Therefore, one effectively will be forced into a 2-component approach, i.e. to render only a tiny portion of the whole galaxy in terms of say 50k stars when the observer is getting really close.

FYI, ... whilst once entered in it, "real" stars are good as well

~50k "real" Celestia stars for each small region of FoV would give a HUGE amount of STC data to read in altogether per galaxy! This is hardly practicable... If instead internal fake stars are used, they can be quickly generated and deleted from memory immediately after the screen-mapped FoV changed into another one (due to a moving observer).

The point is that the mapping of sprite-intensity to star coordinates is different for each small portion of FoV!

Fridger