Celestia Forums

Hello Everyone,

I have done several studies for my boss using Celestia. He continues to question the accuracies of Celestia. I promised him I would look into the accuracy of this great tool. He said that other packages promise 0.2 arc sec accuracies. The tool he is refering to is for aligning theatolites.

My question to everyone is how can we quantify the rendering of the star positions? I think the rendering accuracy is only as good as the input star data. Is the star's RA/Dec very high resolution? I know the C++ class that manipulates the positions are 16 bytes per component. That is very high resolution.

Thanks,
Tim

tec wrote:He said that other packages promise 0.2 arc sec accuracies. The tool he is refering to is for aligning theatolites.

Are you sure your boss knows what he's talking about? . I bet you don't know what a 'theatolite' is?

How about a theodolite which is a surveyor's instrument for measuring horizontal and usually also vertical angles

The rendering of the star positions is perfectly fine within the limitations from OpenGL, of course. The positions are from one of the best scientific catalogs. So nobody can do better than these data.

It's a much more critical issue to benchmark the accuracy of events within our solar system, like e.g. occultation events of the iovian moons that are VERY accurately known.

Here we made many tests with great results.

Bye Fridger

Star positions are actually stored as single precision floating point values, but I don't think that this will be a problem. Single precision floats have a 23 bit mantissa, which gives a precision of about 1x10^-7 (and that doesn't include the extra bit for sign.) Arcsecond precision is only 1.3x10^-6 . . . The star positions are stored in cartesian coordinates, not spherical coordinates, but there's enough extra precision in single precision floating point values that Celestia should still have sub-arcsecond precision.

One thing that could be a problem is the fact that Celestia doesn't account for stellar proper motion. For most stars, this isn't a problem. But there are a few for which the position will be noticeably incorrect. Barnard's Star has the largest proper motion--about 10 arcseconds per year. This is extreme, however. There are less than 10 stars with 5 arcsecond/year proper motions. Is this going to be a problem for you? Here's a list of stars from the HIPPARCOS catalog with high proper motions:

http://www.rssd.esa.int/SA-general/Proj ... le362.html

Adding support for proper motion to Celestia is nontrivial, but I have a plan for how it might be accomplished.

--Chris

chris wrote:Adding support for proper motion to Celestia is nontrivial, but I have a plan for how it might be accomplished.

--Chris

Chris,
Just wondering what impact (if any) this would have on celURL's and Bookmarks around those stars?

Chris,

Celestia already has support for proper motion: it's called EllipticalOrbit

Celestia will happily draw stars in our stellar neighborhood which have orbits defined relative to the galactic center, for example. The units are somewhat awkward (AU), and motion vectors need to be translated into an appropriate coordinate system, but it does work. I've only tested it with gross motions, though. Increased precision might be needed.

That's thinking outside the square ... er...ellipse Selden!

That said we only know the proper motions for a relatively small number of stars in our galaxy (the ones nearby)...

selden wrote:... Increased precision might be needed.

I suspect you might come up against similar accuracy issues as in this example: http://celestiaproject.net/forum/viewtopic.php?p=78873#78873 (albeit on a smaller scale)