Eclipse Finder +++, now what ?

[Calculus]

Congrat again Kendrix, your eclipse finder is an incredible addition to Celestia. I wish It could be extended to any system/body. Chris said the problem was only with the user interface (choose of a planet or moon etc...)

Now Kendrix, are you working on a similar addition ? like
- occulations (when does a moon go into a shadow of a planet or behiond a planet?)
- transits (when does a moon/planet pass in front of another planet/sun?)

Actually, it would be nice to be able to choose the POV for these calculations (earth or another planet or another moon or anywhere).

I suppose, it is very similar to the eclipse finder, and the issue is really the user interface.

Another question would be :
when and what an asteroid/moon/planet can occult during a period of time ? (I mean stars or any other body present in Celestia).

Is this something interesting for other Celestia users or am I the only one (with Kendrix I suppose) ?

[selden]

I have to admit I'd be interested in occultations. E.G. what stars will Cassini see occulted by Saturn's rings.

[Calculus]

May be this one is a little too much!!! I think occultation by saturn is enough, And then you go to the time/location a see directly with your eyes if the star is occulted by the rings.

However, to be certain that no event is forgotten when you do this kind of computation (due to lack of precision of the ephemeris) , astronomers "increase" the actual radius of the planet/moon to calculate the circumstances of the event.

The eclipse finder or the occultation finder shall have an option for computation with greater radius.

[selden]

Paul,

Well, of course the accuracy isn't going to be very good, but I can dream, can't I?

It's challanges like this that help us improve. If NASA can predict the positions of their spacecraft with adequate precision to take pictures of tiny dark asteroids that they couldn't see in advance, the equivalent certainly should be possible in Celestia.

Celestia's spacecraft orbital calculations when derived from the information available from JPL's Horizons ephemeris service is another indication of the difficulty. Celestia's predictions usually differ significantly from values that NASA has published for various events.

Stardust's recent flyby of the asteroid Annefrank is one example. Celestia's predicttion of their separation is off by 2,500 kilometers. This can make quite a large angular difference when trying to do stellar triangulations. Of course, it's also hard to tell where the errors are: in Celestia's calculations or Horizons'. I'm beginning to suspect problems in Horizons' database, myself.

[Calculus]

I agree, let's dream. Isn't it what we are all doing, using Celestia ?
About precision, there might be errors in Horizon, but we certainly know some in Celestia. I Hope Chris will fix them "some day".
About the Eclipse/Occultation/Transit finder, my understanding is that Kendrix uses formulas in Celestia whatever they are. It is just a tool, assuming the positions calculated in Celestia are correct.
But let's hear Kendrix a little on this matter. Please.

[chris]

Paul,

Well, of course the accuracy isn't going to be very good, but I can dream, can't I?

Positions of the major planets are now quite accurate; I've tested them against JPL data, and the errors are sub-arcsecond. Of course, this is still not good enough for something like the Galileo-Amalthea encounter.

For phenomena like occultations, it's necessary to account for aberration. Celestia doesn't do that yet, so that will limit its accuracy.

It's challanges like this that help us improve. If NASA can predict the positions of their spacecraft with adequate precision to take pictures of tiny dark asteroids that they couldn't see in advance, the equivalent certainly should be possible in Celestia.

Well, Celestia does have to do everything in real time, so some compromises may be necessary.

Celestia's spacecraft orbital calculations when derived from the information available from JPL's Horizons ephemeris service is another indication of the difficulty. Celestia's predictions usually differ significantly from values that NASA has published for various events.

It depends . . . Again, planet positions are quire good. Minor planet positions will be inaccurate since they're only computed using ellipses.

Stardust's recent flyby of the asteroid Annefrank is one example. Celestia's predicttion of their separation is off by 2,500 kilometers. This can make quite a large angular difference when trying to do stellar triangulations. Of course, it's also hard to tell where the errors are: in Celestia's calculations or Horizons'. I'm beginning to suspect problems in Horizons' database, myself.

There may be some problems in the Horizons database, but I inclined to think that the problems is with Celestia. Interpolation errors in the spacecraft trajectory could be part of the problem. I'm only doing linear interpolation right now. In future versions of Celestia, I'll implement much better techniques.

--Chris

[selden]

Well, for Stardust's flyby of Annefrank, I used eliptical parameters for Annefrank's orbit as calculated by Horizons for the date of the flyby. And for Cassini's xyz trajectory the samples were about every 15 minutes. The ssc files in the .zip archive on my Web site.

Celestia 1.2.5pre7 shows a closest approach of about 606km at about 2002-11-02 04:44 UT. NASA's published prediction was for a closest approach of 3,300km at 04:50. See http://stardust.jpl.nasa.gov/news/status/021102.html. Celestia shows a 3,000km distance at about 15 minutes before and after closest approach. I guess that's both incredibly good agreement... and disappointing at the same time.

p.s. I just fetched an xyz trjectory for Annefrank with 15 minute sample intrvals for the day of the fllyby. The position differs from the elliptical orbit by one entire kilometer at the time of the flyby. And as much as 20km at other times. I dunno why you seem to think Celestia's elliptical calculations are so bad.

[chris]

Well, for Stardust's flyby of Annefrank, I used eliptical parameters for Annefrank's orbit as calculated by Horizons for the date of the flyby. And for Cassini's xyz trajectory the samples were about every 15 minutes. The ssc files in the .zip archive on my Web site.

Celestia 1.2.5pre7 shows a closest approach of about 606km at about 2002-11-02 04:44 UT. NASA's published prediction was for a closest approach of 3,300km at 04:50. See http://stardust.jpl.nasa.gov/news/status/021102.html. Celestia shows a 3,000km distance at about 15 minutes before and after closest approach. I guess that's both incredibly good agreement... and disappointing at the same time.

Stardust's velocity around the time of the Annefrank encounter was roughly 15 km/s. So it traveled 13,500 in 15 minutes; that's a big distance, but I don't think that the spacecraft's trajectory curved enough to give 2500km of interpolation error. Another possible source of error other than Horizons is Celestia's elliptical orbit calculation. I don't believe that it contains any bugs, but a fixed number of iterations are performed when solving Kepler's equation. I'll try bumping up the number of iterations to see what difference it makes.

p.s. I just fetched an xyz trjectory for Annefrank with 15 minute sample intrvals for the day of the fllyby. The position differs from the elliptical orbit by one entire kilometer at the time of the flyby. And as much as 20km at other times. I dunno why you seem to think Celestia's elliptical calculations are so bad.

I guess it's not the elliptical calculations . . . I never thought they were bad; it's just that ellipses don't approximate orbits that well over long periods of time. How long did it take to get 20km of error?

I'd try getting values from Horizons for the position of Stardust and Annefrank at the time of closest approach. If they're off, then the discrepancy is definitely not Celestia's fault . . . Otherwise, since you've eliminated the elliptical orbit calculation as the source of error, it must be the interpolation . . .

--Chris

[selden]

Chris,

The elliptical orbit parameters for Annefrank that I'm using are supposed to be valid as of 2002-Nov-02 00:00:00.0000 (CT).

Celestia shows a 20km offset beween the elliptical orbit and the xyz trajectory at about +/- 12 hours from that time. Most of the excursions seem to max at about +/- 8 to 15km for closer times.

The errors are on one side of the xyz trajectory before the epoch of the elements and on the other side afterward, both gradually increasing as the time away frrom the epoch increases.

Now I'll go try to persuade Horizons to tell me the time and distance of closest approach...

[selden]

The problem is at Horizons. Or else Celestia can't calculate the distance between two points in a Cartesian coordinate system

It turns out that Horizons will only calculate a closest approach if one of the bodies is one of the 9 major planets or the 4 largest main-belt asteroids.

So I downloaded higher sample rate xyz trajectories: 300 samples for +/- a half hour aound the time of closest approach for both Annefrank and Stardust. Closest approach is shown to be 605.009 km at 2002 11 02 04:43:34.

pffft to Horizons.

I'll see if I can find an address to write to them about this.

sigh.

[selden]

I just heard from one of their engineers. Apparently the Stardust team hasn't provided an updated trajectory since last year. They've taken it out of the database