Scientifically accurate ?

[nrathke]

Hi,

I am not sure which list to post this to. Are the planet orbits scientifically accurate ? I read a post on the Dev list where it seemed to indicate that some parts of Celestia were not scientifically accurate but it was not clear to me which parts it was referring to.

Thanks.

Nick Rathke

[selden]

The orbits of the planets in Celestia are specified using what is known as "VSOP87 Theory". As seen from the Earth, they are accurate to better than one second of arc for about +/- 3000 years from today. Equivalent orbital ephemerides are used for most of the major moons: those in solarsys.ssc which have CustomOrbit specifications.

VSOP87 is a set of polynomials describing the orbits of the major planets. There are over 1000 terms in each series.

See
http://adsabs.harvard.edu/cgi-bin/nph-b ... 202..309B&
and
http://cdsweb.u-strasbg.fr/cgi-bin/Cat?VI/81

Ref:

Planetary theories in rectangular and spherical variables - VSOP 87 solutions
Authors:
Bretagnon, P.; Francou, G.
Affiliation:
AA(Bureau des Longitudes, Paris, France), AB(Bureau des Longitudes, Paris, France)
Journal:
Astronomy and Astrophysics (ISSN 0004-6361), vol. 202, no. 1-2, Aug. 1988, p. 309-315.

If you need more precision, Celestia v1.5.0 also can use the various JPL ephemerides. The Celestia Wikibook describes how to do this. See http://en.wikibooks.org/wiki/Celestia/JPL_Ephemerides

[nrathke]

WOW! that was just the info I was looking for !

Thanks for the fast reply.

[t00fri]

I don't know at what "level of insight" you were asking about Celestia's "scientific orbit accuracy". As Selden wrote, VSOP87 usually gives very accurate performance. Most people who know what VSOP87 really does, also know that the only relatively "weak" orbit in this framework is that of the Moon. For the Moon, there are the following more accurate but also much more CPU-expensive alternatives: VSOP2000 and the numerical integration algorithms from JPL: DE405, DE406 ... In the latter cases one would have to carry along HUGE numerical data files...

DE405 includes both nutations and librations. Then there is of course the "classical" DE200 solution (nutations but not librations) that formed the basis of the Astronomical Almanac since 1984...

Bye Fridger

[BobHegwood]

Good Doctor,

Can the Brain-Dead ask WHY the Moon is different? What makes it
so different from any other orbiting body in the Solar System?

Just curious...

Thanks, Bob

[t00fri]

Good Doctor,

Can the Brain-Dead ask WHY the Moon is different? What makes it
so different from any other orbiting body in the Solar System?

Just curious...

Thanks, Bob

Bob,

first you should recall what VSOP87 qualitatively does. It is some kind of "perturbation expansion", as we say, which successively takes into account the cumulative effects of all sources of a gravitational field (other planets, moons,...) around a given body in form of hundreds of "perturbative" terms. [It's precisely the analog of the so-called Hartree-Fock method in atomic physics.]. These correction terms due to the other bodies are relative to elliptical Keplerian orbits which are assumed as a starting approximation.

The specific problem with the Moon is quite intuitive and more general than VSOP87, actually:

The Moon moves noticeably differently from a simple Keplerian ellipse, because of the competing gravitation of the Earth and the Sun.. So many more correction terms are actually needed. That and many other more technical issues form the basis for why the Moon orbits are difficult to calculate at high accuracy. You will perhaps also recall that the gravitational force is a long-range force (decreasing relatively slowly with distance) and its strength is proportional to the masses of the bodies under consideration. So the Sun is still very competitive at the location of the Moon. And so is the Earth, of course.

Bye Fridger

[selden]

A generic comment:

I've often felt some amusement about the large number of terms used used in the mathematical series to describe orbits. Elliptical orbits around the Sun were supposed to be much simpler than the older orbital descriptions which used epicycles upon epicycles. However, the precision that we need today seems to have resulted in something almost as bad....

[Cham]

A generic comment:

I've often felt some amusement about the large number of terms used used in the mathematical series to describe orbits. Elliptical orbits around the Sun were supposed to be much simpler than the older orbital descriptions which used epicycles upon epicycles. However, the precision that we need today seems to have resulted in something almost as bad....

Well, this is just a manifestation of the extreme precision we're currently able to achieve with measurements. Reality is always much richer than what theory is saying. This is true, also, for the "simple" harmonic motion of a mass suspended to a spring. In this case, the theoretical motion is simply given by a trigonometric formula (x = A sin(wt + phi)), while the true motion is very complex (because of the action of frictional forces, perturbations from the true shape of the spring, etc).

[t00fri]

A generic comment:

I've often felt some amusement about the large number of terms used used in the mathematical series to describe orbits. Elliptical orbits around the Sun were supposed to be much simpler than the older orbital descriptions which used epicycles upon epicycles. However, the precision that we need today seems to have resulted in something almost as bad....

Yes but we must fold in that the accuracy of our admittedly complicated descriptions has increased tremendously compared to the times of epicycles etc.

"Perturbative expansions" are of course a standard tool in theoretical physics and are successfully applied in many different fields. The underlying idea being always that one can identify in the problem a small quantity/parameter, alpha, say, such that a "perturbative expansion" of the exact solution around the limiting case alpha=0 is justified. The resulting algorithmic expressions are often VERY accurate, but usually look quite horribly complex.... that's the price for accuracy.

Bye Fridger

[BobHegwood]

The Moon moves noticeably differently from a simple Keplerian ellipse, because of the competing gravitation of the Earth and the Sun.. So many more correction terms are actually needed. That and many other more technical issues form the basis for why the Moon orbits are difficult to calculate at high accuracy. You will perhaps also recall that the gravitational force is a long-range force (decreasing relatively slowly with distance) and its strength is proportional to the masses of the bodies under consideration. So the Sun is still very competitive at the location of the Moon. And so is the Earth, of course.

Bye Fridger

Thanks very much for the explanation. You probably don't realize this,
but I had never even heard the term "VSOP87" until I saw it used in
Celestia. I actually understood about 1/10 of what you said in the
paragraph above the one quoted here.

At any rate, the explanation WAS much appreciated. Can I make
a point here though? Not EVERYONE who is very much interested
in Celestia has the educational background - or even the BRAINS -
necessary to fully understand the theories and mathematics
behind the program.

As you may have noticed though, I try not to let that bother me.

[chris]

I don't know at what "level of insight" you were asking about Celestia's "scientific orbit accuracy". As Selden wrote, VSOP87 usually gives very accurate performance. Most people who know what VSOP87 really does, also know that the only relatively "weak" orbit in this framework is that of the Moon. For the Moon, there are the following more accurate but also much more CPU-expensive alternatives: VSOP2000 and the numerical integration algorithms from JPL: DE405, DE406 ... In the latter cases one would have to carry along HUGE numerical data files...

DE405 includes both nutations and librations. Then there is of course the "classical" DE200 solution (nutations but not librations) that formed the basis of the Astronomical Almanac since 1984...

You can use JPL ephemerides with Celestia if you need more accuracy than VSOP87. The instructions for doing so are here:

http://en.wikibooks.org/wiki/Celestia/JPL_Ephemerides

--Chris

[t00fri]

You can use JPL ephemerides with Celestia if you need more accuracy than VSOP87. The instructions for doing so are here:

http://en.wikibooks.org/wiki/Celestia/JPL_Ephemerides

--Chris

What's the performance ratio between using the DE 405 LUT's and VSOP87 in Celestia?

Bye Fridger

[chris]

You can use JPL ephemerides with Celestia if you need more accuracy than VSOP87. The instructions for doing so are here:

http://en.wikibooks.org/wiki/Celestia/JPL_Ephemerides

--Chris

What's the performance ratio between using the DE 405 LUT's and VSOP87 in Celestia?

Bye Fridger

DE405 should be considerably faster, though I haven't actually measured the difference. The only drawback is the size of the data files. DE406 is better, but you still need a lot of data.

--Chris

[t00fri]

You can use JPL ephemerides with Celestia if you need more accuracy than VSOP87. The instructions for doing so are here:

http://en.wikibooks.org/wiki/Celestia/JPL_Ephemerides

--Chris

What's the performance ratio between using the DE 405 LUT's and VSOP87 in Celestia?

Bye Fridger

DE405 should be considerably faster, though I haven't actually measured the difference. The only drawback is the size of the data files. DE406 is better, but you still need a lot of data.

--Chris

What format are these DE 40x files? Ascii or binary? We could always use gzipped versions and decompress on the fly... May be worth checking what the compression ratio will be.

Bye Fridger

[ElChristou]

DE405 should be considerably faster, though I haven't actually measured the difference. The only drawback is the size of the data files. DE406 is better, but you still need a lot of data.

In the case one day we manage to build a Highres official package for download, what about including new data set next to the maps and models?

[t00fri]

DE405 should be considerably faster, though I haven't actually measured the difference. The only drawback is the size of the data files. DE406 is better, but you still need a lot of data.

In the case one day we manage to build a Highres official package for download, what about including new data set next to the maps and models?

Would make up for a keen 20 GB download package

Bye Fridger

[cartrite]

You can use JPL ephemerides with Celestia if you need more accuracy than VSOP87. The instructions for doing so are here:

http://en.wikibooks.org/wiki/Celestia/JPL_Ephemerides

--Chris

What's the performance ratio between using the DE 405 LUT's and VSOP87 in Celestia?

Bye Fridger

DE405 should be considerably faster, though I haven't actually measured the difference. The only drawback is the size of the data files. DE406 is better, but you still need a lot of data.

--Chris

What format are these DE 40x files? Ascii or binary? We could always use gzipped versions and decompress on the fly... May be worth checking what the compression ratio will be.

Bye Fridger

I have the file de405.bsp in my ISIS data folder and it is mostly binary. There is ascii text embedded though. I think there is a tool in the spice toolkit that can strip out the text, but I don't think that would do much good? The file is 10.4 mb and I can gzip it to 9.9. I think that is all you'll get from stripping the ascii text from the file too. Not sure.
cartrite

[t00fri]

Ah thanks, cartrite.

That means that they are working already with a quite economical binary format. On the other hand every decent texture is 10 MB these days

It might be worth examining these DE 405 orbits a bit closer...

Bye Fridger

[ElChristou]

if the improvement is ggod enough, 10 mo are not THAT bad...

[chris]

There are different JPL ephemerides. I think that the one of most interest to Celestia users is DE406, not DE405. The DE406 ephemeris covers the time span from years -3001 to +3000, while DE405 spans 1600 to 2200. Each 300 year block of DE406 is about 10 MB, so 200 MB for the whole 6000 years covered. DE405 runs to 4.7 MB for each 50 years, so it's roughly three times as large as DE406 for the equivalent timespan. DE405 also includes nutations and librations; Celestia doesn't make use of these yet.

Regarding the difference in accuracy between DE405 and DE406, the JPL documentations says this:

This is the same ephemeris as DE405, though the accuracy of the
interpolating polynomials has been lessened (interpolation on the
64-day mesh points remains exact, however). For DE406/LE406, the
interpolating accuracy is no worse than 25 meters for any planet and
no worse than 1 meter for the moon.

Good enough for most Celestia users, I think . . .

--Chris