Celestia Forums

Hello

I was surprised by the big difference in the positions of the Martians moons in Celestia 1.5.1 and in Celestia 1.6.0.
To see witch was the right one I decided to compare them with this tool from skyandTelescope.com:
http://www.skyandtelescope.com/observin ... ndex.html#
That tool gives the (hopefully) exact position of the two satellites for Dec 2007 and Janv. 2008

I used a little .ssc file (I join this file with the post) to add two versions of Phobos and Deimos, with position calculated with the old fashion of celestia 1.5.1. So I can view both old and new positions in celestia 1.6.0.
I called these moons Demo and FauxBoss.


I put myself in the center of the Earth (alt -6000km)
and Here is the result for December 11 2007 at 12h00



And the cel:: link to it :
cel://SyncOrbit/Sol:Earth/2007-12-11T12:01:05.18331?x=yHPfM+CZ/f///////////w&y=mSiWldmg/////////////w&z=GgYmQxEE/////////////w&ow=0.573032&ox=-0.157138&oy=-0.787233&oz=0.164944&track=Sol:Mars&select=Sol:Mars2:Demo&fov=0.0203328&ts=1&ltd=0&p=1&rf=75617687&lm=2052&tsrc=0&ver=3


Comparing this with the position given by the Sky and Telescope tool (attention to click on the ‘Direct View’ of the telescope), it appears that the winners are Demo and FauxBoss while Deimos and Phobos are in the wrong place.

Therefore when the rotation model of Mars is updated, the orbits of its satellites should be redefined accordingly.

Maybe there is the same problem with spacecrafts orbiting around the Earth, but I am unable to verify that.

What a horror! I can definitely confirm your findings.

While I don't trust the S&T tool (containing NO statements about underlying astro-math and accuracy) I do trust XEphem,

http://www.clearskyinstitute.com/xephem/xephem.html

where I was involved for 11 years .

The positions of Mars, Deimos and Phobos as seen from Earth in Celestia 1.5.1 agree perfectly with XEphem's Mars moon views (in 2 othorgonal projections) for Dec 11 2007, 12:00:00 UT. See also the above correct image with Mars2/Demo/FauxBoss.

However, the positions of Deimos and also Phobos are severely incorrect in the latest Celestia 1.6.0SVN.

While I have spent many hours of carefully testing the moon orbits of all planets against highly accurate events (including mutual eclipses) a number of years ago, someone among us devs has implemented recently orbit changes without doing careful checks afterwards!

What is all this "3d splendor " of Celestia worth, if possibly most of our satellite orbits are now in error! And that shortly before the release of 1.6.0...

Fridger

The parameters of the Phobos orbit don't appear to have been changed. I suspect the problems arise because Mars now has a CustomRotation and BodyFrame defined, which is messing up the interpretation of the satellite orbital elements.

ajtribick wrote:The parameters of the Phobos orbit don't appear to have been changed. I suspect the problems arise because Mars now has a CustomRotation and BodyFrame defined, which is messing up the interpretation of the satellite orbital elements.

In my solarsys.ssc Mars, Phobos and Deimos orbits have ALL changed, along with many other celestial bodies, the orbit accuracy of which now needs to be tediously asserted!

Here is Phobos: in 1.6.0SVN

BodyFrame { EquatorJ2000 {} }
CustomRotation "iau-phobos"

# Overriden by CustomRotation
# UniformRotation
# {
# Inclination 0.0
# MeridianAngle 354.2
# }

In 1.5.1 Windows XP I have instead for Phobos::
UniformRotation
{
Inclination 0.0
MeridianAngle 354.2
}

For Deimos in 1.6.0SVN:

BodyFrame { EquatorJ2000 {} }
CustomRotation "iau-deimos"

# Overriden by CustomRotation
# UniformRotation
# {
# Inclination 0.9
# AscendingNode 183.0
# MeridianAngle 214.7
# }

in 1.5.1 Windows XP I have instead:

UniformRotation
{
Inclination 0.9
AscendingNode 183.0
MeridianAngle 214.7
}

So why did you claim that Phobos was unchanged?

Fridger

t00fri wrote:So why did you claim that Phobos was unchanged?

I said that the orbital parameters have not been changed. The parameters you have quoted are all to do with the rotational parameters of the object. This shouldn't have an effect on where the object is in its orbit!

E.g. the .ssc section for Phobos from the version of solarsys.ssc in v1.5.1 as found on the SVN here is


The definition of Phobos in the trunk is:



The orbital parameters (in the EllipticalOrbit block) are the same, yet the positions of the moons are different in v1.5.1. As far as I can tell this is because the reference frame for the moons has changed due to the change in the definition of the rotational frame for Mars, which is used as the default orbital frame for Mars' satellites.

Here's the difference between the definitions of Mars:

Here's v1.5.1


Here's the current trunk version


As you can see, Mars now has its body frame set to EquatorJ2000, which I suspect results in a change to the default orbital frame of the satellites, hence without changing the numeric values for the orbital parameters, the positions of Phobos and Deimos are different.

Oh, sorry, yes I misread the subtle part of your statement. Still,... ALL that matters is that one now observes a LARGE discrepancy in the orbital locations.

I have seen NO displays of critical orbit tests, after these massive modifications were done...

This shouldn't have an effect on where the object is in its orbit!

Without seing explicit, quantitative checks I do NOT conclude that such "harmless" new statements like

BodyFrame { EquatorJ2000 {} }
CustomRotation "iau-phobos"

indeed do exactly what they are supposed to do!

As a good illustration, just have a quick look here

viewtopic.php?f=3&t=1506&hilit=mutual+events

demonstrating how a stupid little bug (that I located) could mess up all Gallilean orbits in 2002

I am currently testing again the very sensitive "mutual events" (eclipses of Gallilean moons, special configurations of Saturnian moons etc) . I have observed already quite significant changes in the event times in the two versions.

Fridger

t00fri wrote:Oh, sorry, yes I misread the subtle part of your statement. Still,... ALL that matters is that one now observes a LARGE discrepancy in the orbital locations.

Pardon the query from the Brain-Dead, but can one get around these discrepancies by
using your code from the previous SSC posts here?

Thanks, Bob

BobHegwood wrote:

t00fri wrote:Oh, sorry, yes I misread the subtle part of your statement. Still,... ALL that matters is that one now observes a LARGE discrepancy in the orbital locations.

Pardon the query from the Brain-Dead, but can one get around these discrepancies by
using your code from the previous SSC posts here?

Thanks, Bob

Yes, Bob, but we GOT to find what is wrong here. Many celestial bodies might be affected.

Fridger

Here's a test .ssc which can be placed in the extras folder

This should be used with the trunk version of solarsys.ssc. "Mars B" is defined with the Mars rotation parameters from v1.5.1, and "Phobos B" is placed in orbit around "Mars B", but has the exact same ssc definition as Phobos does in the trunk. "Phobos C" meanwhile is defined with the v1.5.1 Phobos rotation parameters, and is placed around Mars as defined in the trunk.

Since "Phobos C" appears in the same place as the incorrect trunk version of Phobos, despite using the Phobos parameters that worked in v1.5.1, the problem is clearly with the change in Mars's rotation frame being propagated to the orbit frame of the satellites.

t00fri wrote:Yes, Bob, but we GOT to find what is wrong here. Many celestial bodies might be affected.

Fridger

I understand...
Many thanks for the explanation and Good Luck.

I think this illustrates the dangers of using the parent body's equatorial frame to define satellite orbits - if you redefine the rotation of the planet, you risk breaking lots of satellite orbits.

ajtribick wrote:I think this illustrates the dangers of using the parent body's equatorial frame to define satellite orbits - if you redefine the rotation of the planet, you risk breaking lots of satellite orbits.

TRUE, indeed. I have never been a friend of using frames in this form ...

Here is at least one test with 1.6.0SVN that has survived very well:
--------------------------------------------------------------------------------------

The precise timing of the annular 60% shadow of Europa on Io (event 2e1A)
on Dec 27, 2002, start 7:20 UT end 7:30 UT

Here is what I get spot on, at 7:23 UT with LT (LIght travel delay) activated


You must realize that Europa is far away from Io at this moment, hence the directions must be perfect as well. Moreover, if LT was not activated the eclipse had taken place MUCH earlier at 6:47 UT instead...

Fridger

Looks like Mercury, Venus, Earth and Mars have had updates to their rotation definitions post-1.5.1. Having tested the solar eclipse of 11 Aug 1999, I suspect objects on CustomOrbits are ok, which means the problems are mainly going to be the minor moons. Probably should check the elements of the various Earth-orbiting satellites if they are defined by EllipticalOrbits. Judging by the comments in the .ssc file, Saturn has had an update to its rotation parameters at some point (this happened before 1.5.1) so probably worth checking those too.

ajtribick wrote:Looks like Mercury, Venus, Earth and Mars have had updates to their rotation definitions post-1.5.1. Having tested the solar eclipse of 11 Aug 1999, I suspect objects on CustomOrbits are ok, which means the problems are mainly going to be the minor moons. Probably should check the elements of the various Earth-orbiting satellites if they are defined by EllipticalOrbits. Judging by the comments in the .ssc file, Saturn has had an update to its rotation parameters at some point (this happened before 1.5.1) so probably worth checking those too.

Did anybody perform a PRECISION test of the position of our Moon after the changes??? I won't do it AGAIN . "Fortunately", Mercury and Venus don't have satellites

Fridger

t00fri wrote:Did anybody perform a PRECISION test of the position of our Moon after the changes??? I won't do it AGAIN

It's all very well saying you won't do the tests again, but at some point such precision tests need to be done, so some instructions on how to do a good precision test would be nice.

ajtribick wrote:

t00fri wrote:Did anybody perform a PRECISION test of the position of our Moon after the changes??? I won't do it AGAIN

It's all very well saying you won't do the tests again, but at some point such precision tests need to be done, so some instructions on how to do a good precision test would be nice.

Well I did VERY many 5 - 6 years ago and found several bugs. I even did very fancy tests for Pluto and Charon. Just use the search engine if you are interested.

+++++++++++++++
If I had changed the orbit code I would also feel obliged to re-assert the positional accuracy "before and after"! This has apparently been omitted, such that now it is much more work.
+++++++++++++++

Here is a URL from the institute of Jean Meeus with many mutual event tabulations (of highest precision!) that are free of charge. Unfortunately, the data by Jean Meeus in S&T 2002, p. 100 are not free.

http://www.imcce.fr/page.php?nav=en/eph ... nomena.php

Fridger

For the Moon it seem there is no difference in position from the surface between v151 and 160. But there is one if we aim at a particulary point.

Moreover there is a difference whith the version 1.4.1 in witch the Earth obliquity was -23.45 instead of -23.4392911 in version 1.5.1

To test all that I begin to place myself with the version 1.4.1 at 10.000km above the Moon surface and Hipparchus.
Then I copy the position in freeflight, so I record an absolute position in space.
Here is the URL
cel://Freeflight/2008-11-06T08:13:00.00001?x=4GJHrlQoTe/IDA&y=nT/9tlbH+P///////////w&z=G6gz/7bpgR31/////////w&ow=0.477556&ox=-0.021268&oy=0.877474&oz=-0.039079&fov=33.151928&ts=0.000000&ltd=0&rf=40435&lm=49286

Then I open that URL with Celestia 1.5.1, (applying corrections depending on the new LY_PER_km and the modification of the centre of the Universe).

I found that the distance from the moon is now 6.330 km instead of 10.000 km
and the distance from Hipparchus is now 51.263 km instead of 10.000 km

I record again this place. Here is the link :
cel://Freeflight/2008-11-06T08:13:00.00001?x=xwczAE0/MD4L&y=M+vZ5X0O1P///////////w&z=LOLbVdrl6wj1/////////w&ow=0.477556&ox=-0.021268&oy=0.877474&oz=-0.039079&fov=33.151928&ts=0.000000&ltd=0&p=0&rf=40435&lm=49286&ver=2

Then opening this link with celestia 1.6.0 leads to:

dist from the moon = 6.3303 km (one more digit only on the display). We may considere it is the same than above
dist from Hiparchus = 43.405 km. I considere that normal since the rotation model of the Moon is also changed

It is an empiric method but several tests show that there is no significant difference between v1.5.1 an v1.6.0.

But there is one between 1.4.1 and the following versions. I am unable to say if it was a bug of the previous versions corrected in v151 and 160 or if it is a bug in 1.5.1 inherited in 1.6.0.

I don't know if this is helpful or not...

A very useful set of tests are occultations of planets and the Moon. These partly grazing occultations have been VERy precisely timed, can be looked up at the official places and serve as excellent precision tests/constraints e.g. for the Moon.

Here is a nice example photo:



++++++++++++++++++++++++++++++++++++
With Celestia 1.6.0SVN we see that the Moon orbit seems pretty good (after a second try of mine )
The location is York/UK where the reappearance of Saturn was at March 2, 2007; 2:51 UT: Despite the good agreement, I still think far too little time is spent on precision tests of Celestia!
++++++++++++++++++++++++++++++++++++


I performed many such tests with Celestia > 6 years ago.
It's all in this forum.

Here is another photo of this event, with the full grazing path beautifully apparent


Here is a page with all details and the times of disappearance and reappearance of Saturn for various locations in England for this event.

http://www.popastro.com/sections/occ/satocc_mar2007.htm

Fridger

Here are the detailed data for another

Occultation of Saturn by the Moon: 16 April 2002,
Glasgow/UK: DD 20h 47m 12s UT RB 21h 25m 18s UT
DD= Disappearance at Moon's Dark limb
RB= Reappearance at Moon's Bright limb

http://www.popastro.com/sections/occ/satapr.htm

This one fits actually to better than 1 minute (from Glasgow/UK)


Unfortunately I did not manage to find the crater names in the reference photo on the occultation page.



Fridger

Well the Moon's orbit is explicitly defined to be in the EquatorJ2000 frame and the Moon is located in the same place under the 1.5.1 and 1.6.0 definitions of the Earth's rotation, which implies that in addition to the problems with the Keplerian orbits, the CustomOrbit needs to be looked at.

In summary, aaaaargh.