Eliminated Bug in Galilean Orbits=> Spectacular Accuracy!

[t00fri]

Chris:

I have carefully measured (by blowing-up + overlay) the positions of the Saturn moons corresponding to the HST photo above, after your recent CVS modifications meant to cure the B1950/J2000 issue.

Unfortunately, the agreement is significantly worse now for Rhea, Tethys and Dione. Rhea is too far out in image 3 (bottom above) and not anymore in the ring plane. The other two moons (top image) are too close, instead.

So far the best agreement, arises in the "mixed" parameter setup as of yesterday evening (Grant's EquatorAscendingNode with everything else like yesterday).

Bye Fridger

[chris]

Chris:

I have carefully measured (by blowing-up + overlay) the positions of the Saturn moons corresponding to the HST photo above, after your recent CVS modifications meant to cure the B1950/J2000 issue.

Unfortunately, the agreement is significantly worse now for Rhea, Tethys and Dione. Rhea is too far out in image 3 (bottom above) and not anymore in the ring plane. The other two moons (top image) are too close, instead.

So far the best agreement, arises in the "mixed" parameter setup as of yesterday evening (Grant's EquatorAscendingNode with everything else like yesterday).

You're right--the "mixed" parameters do seem to work better. I'm still investigating . . . I did discover an apparent sign error with the inclinations of the Saturnian moons. After fixing it, I was able to view eclipses of Iapetus on the correct dates.

--Chris

[t00fri]

You're right--the "mixed" parameters do seem to work better. I'm still investigating . . . I did discover an apparent sign error with the inclinations of the Saturnian moons. After fixing it, I was able to view eclipses of Iapetus on the correct dates.

--Chris

Interesting! Where is the sign error? I am tracking various potential bugs in the satellite code as well. One suspicion is again that some terms use degree units and others (to be added together) radians...

From my extensive comparisons of mutual saturnian moon events, I am meanwhile convinced that there are still bugs in the code.

Bye Fridger

PS: I have coded a lengthy patch solving almost completely the "selection problem at small FoV" I reported in the developer mail...

The trick was to use instead of

float cosAngle = starDir * pickray

the much more stable

double sinAngle2 = sqrt ((starDir-pickRay)* (starDir - pickRay))/2.0,

with both starDir, pickRay being unit vectors, of course.

While for small angles of the arc sec range, numerically cosAngle == 1 in /float/,
2 sinAngle2 \approx Angle with all needed digits under control!

Bye Fridger

[chris]

You're right--the "mixed" parameters do seem to work better. I'm still investigating . . . I did discover an apparent sign error with the inclinations of the Saturnian moons. After fixing it, I was able to view eclipses of Iapetus on the correct dates.

--Chris

Interesting! Where is the sign error? I am tracking various potential bugs in the satellite code as well. One suspicion is again that some terms use degree units and others (to be added together) radians...

The sign error is in SaturnMoonPosition() . . . I believe that this line:

Code: Select all

gam = degToRad(gam);


should actually be:

Code: Select all

gam = -degToRad(gam);


My testing so far has born this out, most convincingly in the above mentioned eclipses of Iapetus.

From my extensive comparisons of mutual saturnian moon events, I am meanwhile convinced that there are still bugs in the code.

I've been testing mutual phenomena of the satellites and, more recently, comparing the moon positions against Horizons data. I find that using sampled orbits derived from Horizons is the best way to debug Celestia's orbital calculations. I've been looking for more degrees vs radians bugs, but haven't noticed anything so far . . . In my opinion, the most likely culprit is still the B1950 coordinate system used for the custom orbits of Saturn's satellites.

--Chris

[granthutchison]

So far the best agreement, arises in the "mixed" parameter setup as of yesterday evening (Grant's EquatorAscendingNode with everything else like yesterday).

Yes, I'm seeing a lot of very good alignments at the correct times for occultations, but with one satellite stacked above the other. The impression is that the whole satellite system is tilted by a small amount around an axis in the plane of Saturn's equator but 90 degrees away from my EquatorAscendingNode.
However there are other occasions on which the alignment is mistimed by 20 or 30 minutes, indicating some kind of persisting, intermittent longitude error. So far I haven't been able to nail this to a particular satellite or satellites.

Grant

[t00fri]

So far the best agreement, arises in the "mixed" parameter setup as of yesterday evening (Grant's EquatorAscendingNode with everything else like yesterday).

Yes, I'm seeing a lot of very good alignments at the correct times for occultations, but with one satellite stacked above the other. The impression is that the whole satellite system is tilted by a small amount around an axis in the plane of Saturn's equator but 90 degrees away from my EquatorAscendingNode.
However there are other occasions on which the alignment is mistimed by 20 or 30 minutes, indicating some kind of persisting, intermittent longitude error. So far I haven't been able to nail this to a particular satellite or satellites.

Grant

Using "mixed" configuration as defined above /plus/ Chris sign change in gam (see above), two of my problematic occultation events now come together nicely!

1) Enceladus /totally/ occulting Mimas (2OCC1T): 08/16/1995:
--------------------------------
predicted (on earth) to occur at 3:46:16 UT. Horizons gives LT=72.6269' for that date, hence the event should occur at 2:33:38.3 Celestia UT. It actually occurs /totally/ at 2:33:36 (event midtime)!! So this is perfect.

2) Rhea /totally/ occulting Tethys (5OCC3T): 05/04/1995:
--------------------------------
Predicted to occur (on earth) at 16:12:27 with LT=85.113' for that date. This means 14:47:20 Celestia UT predicted. In Celestia, it occurs, however, at 14:52:40, /totally/ though but quite a bit later.

Still, this is by far the best I had so far...

Bye Fridger

[t00fri]

Here is the great challenge:

3ECL2T, a /total/ eclipse of Encelatus by Tethis, 07/22/1995, 16:01:15.34 Celestia UT (LT=74.911').

The above setup sees NOTHING...

Bye Fridger