Celestia Forums

Hi!

I think that you could do the celestia galaxy not with Billboarding but with actual stars, but you would have to find a possibility of "sdding up" light form distant stars, so that we have a nebula-like view.

I just tested approximately how many stars celestia can handle at the same time... (on screen)

I have not finished yet, but as i iviewed this it was close to the limit... :-)
(Half a million Stars, some invisible)



(Coded a semirandom star distributer, based on Chris' code for the Galaxy display. Thanx Chris!! :-)

Sirius

Sirius wrote:(Coded a semirandom star distributer, based on Chris' code for the Galaxy display. Thanx Chris!! :-)

It looks great! It would be even better if you could vary the distribution of spectral classes to put Pop II stars in the nucleus and Pop I in the arms.

Also, while the current ability to compute galaxy shapes based on classification is OK, and probably vital if we want to have a nice big galaxy database, it would be good to have the ability to go beyond that and use some sort of more accurate shape data, perhaps derived as a best guess from photos. Even the shape of our own galaxy (while hard to gauge since we're inside) is known to somewhat greater accuracy than this symmetrical two-armed shape.

Impressive, how fast is your computer?

Wow...If you add Chris's galactic nebula formation model that would look even more realistic...

I wonder if Chris will let you add this with the upcoming ver 1.2.5!!!

Do the Sombrero Galaxy next well if you want

I thought I had already replied, but obviously the post got lost...

I am working at an advanced semirandom star generator, but for different stars I need the codes for the different star types.
In the last compilation i have just taken "sun-type" for all of them.

@Chris: Could you just mail me a short description of the different star types? It would be VERY helpful .

Sirius

P.S.: An unusual bug (in my prog): http://www.lanbada.de/celestia/halfgal.jpg

That clipping is probably the same thing that happens with the stars that are greater than 6000 ly from Sol...They seem to disappear when facing outwards from the center...

That clipping was due to a "square" ( ^2 ) in the translation code, I adjusted it with adding 180° to the rectascension when the z-coordinate is negative :-)
I just added this picture for fun

Sirius

Sirius wrote:That clipping was due to a "square" ( ^2 ) in the translation code, I adjusted it with adding 180° to the rectascension when the z-coordinate is negative
I just added this picture for fun

Sirius

When are you releasing this addon/hack....I would like to try it out...

The improvement I would like to see most is more deep space objects.

I can see that some people don't want imaginary stuff in celestia, but it'd be nice to have an option to generate random galaxies containing random stars. Of course it would be stupid to render all the stars in all the galaxies at the same time. When you get near a galaxy, it can be rendered as a smudge, as you get closer it can render a smudge + some random stars, as you get closer render more stars, etc.

It would be nice to have randomly generated stars (and even planets!) in well known galaxies, like andromeda. There are pretty good theories about the distribution of galaxies in the universe and these could be used to generate random galaxies.

It will be ok if the random number generator happens to produce different stars each time you visit a galaxy. However, it would be very cool if it didn't, for instance it could use some kind of fractal function.

John Henckel -- johnh@mathengine.com

The random number generator can be initialized with the same number (you could for example derive it from the name of the Galaxy)

but the problem with large galaxies is that for example our galaxy (the Milky Way) has 10 000 000 000 (in words: 10 Billion) stars. Celestia is VERY memory-saving with star data (i had a galaxy with 5 Million stars running on a machine with 256 MB RAM) but the actual number of stars is 2000 times that Value!!! To run that smoothly, It would need a machine with at least 256 GIGABYTE RAM.

Though, we could add a random star generator to the celestia code, but there I would need some more time to work on...

Then we could *simulate* 5 Billion stars, though we haven' the memory to put them in the database (the 1 Million star database was 120 MB)

Could do that on the weekend...

@Chris: Could you just tell me the startype codes, that would be very helpful!!

Sirius

Sirius wrote:The random number generator can be initialized with the same number (you could for example derive it from the name of the Galaxy)

but the problem with large galaxies is that for example our galaxy (the Milky Way) has 10 000 000 000 (in words: 10 Billion) stars. Celestia is VERY memory-saving with star data (i had a galaxy with 5 Million stars running on a machine with 256 MB RAM) but the actual number of stars is 2000 times that Value!!! To run that smoothly, It would need a machine with at least 256 GIGABYTE RAM.

It's very cool that you got a 5 million star catalog working--how fast did Celestia run? The octree visibility culling algorithm used by Celestia should scale very well, but I haven't tried it myself.

Sirius wrote:Though, we could add a random star generator to the celestia code, but there I would need some more time to work on...

I've got some ideas on how to do this that would integrate well with the octree structure.

Sirius wrote:@Chris: Could you just tell me the startype codes, that would be very helpful!!

Are you asking about the spectral and luminosity classes? Like G2 V, K5 II, etc.?

--Chris

It's very cool that you got a 5 million star catalog working--how fast did Celestia run? The octree visibility culling algorithm used by Celestia should scale very well, but I haven't tried it myself.

It run at (approximately) 10-2 fps (depended on the amount of stars visible)
I think it could run much faster when I introduce more different star sizes (result: more culling, less visible stars)

If I have understood the concept of an octree, it would be possible to calculate for each of the "branches" a brighness value, so that it could be rendered as a nebulous object from large distances.

btw, all the screenshots I made were taken with a very wide fov, so that i could zoom in to the galaxy close enough that the stras in the back didn't disappear. If you watch these stars on normal settings from some thousand light years, you see absolutely nothing.

The "color" calculated from the star types and brightness in an octree branch could be a soultion, but as I said, I need to familiarize myself with the code to see if that's possible.

@Chris: Yes, I meant that :-) In the catalog there are only obscure numbers around 800 or 1400, (showed my Hex-Editor)

@chris (2nd time): we could use the galaxy definition file for random star distribution, put it in the code and make it switchable so that users can switch the irreal stars off.
You'd have to add an algorithm, though, that removes all irreal stars visible from earth (we have stars there already :-)

I'll come back Sunday evening (Middle European Time :-)

Sirius

Well, lets just wait for 10 years until this is possible on every new computer. =P

But, we could just put up with the frame rates until then(or get an SGI supercomputer)

Mikeydude750 wrote:Well, lets just wait for 10 years until this is possible on every new computer. =P

But, we could just put up with the frame rates until then(or get an SGI supercomputer)

Bah, with the proper coding Chris can add every damn thing in the known universe without bogging down the computer TOO bad...using line of sight at a limited visible range...those objects not visible will not be rendered...much like ray-casting done in shooter games like Quake etc...
Just gonna be several CD's and tons of GB of HD space needed...By then we will have a DVD version of this software I suspect...since all the info might fit on a DVD sized CD

Dont put it past him...I'll bet in a year this will be the VR of our universe we've always wanted

The solution to all the memory and processor problems is adaptive detail. When you're at 1 ly from the sun, you don't see the earth, so you don't need it to be present to render.

When zooming out to see the entire galaxy, you'll only see the brightest stars, and not all other stuff that would be invisible. That's what should not be loaded in memory.

Problem is that you have to have at least a list and positions of the Objects im memory to calculate a line of sight.
But perhaps Chris could enhance his Octree system so that ist caches unused branches on the hard drive, although that would be slowwww