Celestia Forums

Hello !

I have taken all remarks in account to generate a new stars database for Celestia

Link http://perso.wanadoo.fr/celestia.stars/ to download the database.

What's new in V1.2 ?

- Chi2 Ori is back among other Hipparcos stars that were rejected in the release 1.1, due to their strange parallax values (negative or lesser than 0.0002”)

I applied the same procedure as I did on Hipparcos stars with negative parallax values, to Hipparcos stars with parallax values lesser than 0.0002” to have a better distance estimation for them. Thus 4356 among 5064 Hipparcos stars with bad parallax values have been corrected.

- The database should be more precise for stars that lie within 20 ly from Sun. But of course, there are still new Tycho stars in this range, accepted with good parallax values (relative error < 25%)

- Rigel Kentaurus B is back. It was rejected in V1.1 and replaced by HIP 33626113 due to big error values in RA and declination.


Best regards to all of you

Pascal

Wow! Rigel's hard at work! Great job, thanks!

Off to get it!!! Yeah!
Great work.
And the server is great. It was a one minute download for me :) 250kb/s

The server hasn't been very nice to me, only 2.9 kb/s (although i do only have a dialup)


Mike M.

Last night, I coded up support for Tycho catalog numbers, so this extended star database will work nicely with the next prerelease . . . Should be ready in a few more hours . . .

--Chris

Chris,

is the stars.dat format the same in the next version?

dtessman wrote:Chris,

is the stars.dat format the same in the next version?

Yes, it's the same format as before . . . any suggestions on improving it? I do plan on keeping the primary star catalog file format binary, both for compactness and to keep loading fast.

--Chris

chris wrote:Last night, I coded up support for Tycho catalog numbers, so this extended star database will work nicely with the next prerelease . . . Should be ready in a few more hours . . .

--Chris

Clicking and displaying the catalog numbers does not reliably work anymore under Linux starting 1.2.5 CVS at about pre1 level. It works perfectly including HIP x xxx xxx xxx from Tycho under 1.2.4. and at the later CVS levels several weeks back (including my own checkins), using identical star data (Rigel's stardb-1.1). I tried to locate the problem in the code yesterday, but without much success yet. More recent changes are mainly comets, texture stuff and|or mouse action code?


Bye Fridger

t00fri wrote:Clicking and displaying the catalog numbers does not reliably work anymore under Linux starting 1.2.5 CVS at about pre1 level. It works perfectly including HIP x xxx xxx xxx from Tycho under 1.2.4. and at the later CVS levels several weeks back (including my own checkins), using identical star data (Rigel's stardb-1.1). I tried to locate the problem in the code yesterday, but without much success yet. More recent changes are mainly comets, texture stuff and|or mouse action code?

Good to see you posting here again, Fridger!

Anyhow, I'm not sure what's going on here . . . The code for displaying catalog numbers is the same for Linux and Windows. It's possible that there was some small change that I forgot to check in. In any case, I'll rebuild on Linux tonight and see if I can figure out what's happening.

--Chris

chris wrote:

t00fri wrote:Clicking and displaying the catalog numbers does not reliably work anymore under Linux starting 1.2.5 CVS at about pre1 level. It works perfectly including HIP x xxx xxx xxx from Tycho under 1.2.4. and at the later CVS levels several weeks back (including my own checkins), using identical star data (Rigel's stardb-1.1). I tried to locate the problem in the code yesterday, but without much success yet. More recent changes are mainly comets, texture stuff and|or mouse action code?

Good to see you posting here again, Fridger!

Anyhow, I'm not sure what's going on here . . . The code for displaying catalog numbers is the same for Linux and Windows. It's possible that there was some small change that I forgot to check in. In any case, I'll rebuild on Linux tonight and see if I can figure out what's happening.

--Chris

;-) Well, whenever there is some spare time...

Anyhow postings seem to get much faster through to you than emails. There is another email bug report on separate dds bumpmap files that seems to have gone straight to Nirvana;-)...

I have just inserted your most recent CVS TYC changes into stardb.cpp and celestiacore.cpp of the final 1.2.4 version and the click&display of TYC stars works beautifully.

So something is definitely broken between CVS June 24 and now as to the Linux display.

Bye Fridger

Hi Rigel, it's great that you've released the Tycho data! I've been busy with other things and haven't had time to check out your compilation yet, but I see that you already has a third version out. In version 1.1 you wrote:

Rigel wrote: All stars of parallax values lesser than 0.0002" (d > 5000 pc) have been rejected, either from the Hipparcos or from the Tycho catalog, because their parallaxes are undoubtedly false.

Is this true also for v 1.2, and if so, why don't you just assume these stars to be class III sources (and ignore the parallax) as we discussed before? As an example, a B=11, V=10 star of class V would have a parallax of ~0.025" while if it was a class III source it would have a parallax of ~0.0015". Thus if the measured parallax is close to zero or negative, and the error of the parallax is "significantly" smaller than 0.025", you can be pretty sure that the source is class III and not class V.

Rigel wrote: All Tycho stars with strange B-V (< -0.5 or > 2.0) color index values were rejected.

Except those few with acceptable parallaxes, I assume?

Great work, and good luck with your students..!

/Alexis

nice nice nice work.
Thnaks very much but...

I dunno what happens, but when I dowonload it it stops download @ 99%
seems that the last 23Kb are missing?

TK274, I think it is my provider. It is a personal web page, so there's not much support for it. Chris will maybe mirror the database on his site.


Alexis,

in V1.2 I've put Hipparcos stars with a parallax < 0.0002" to this limit value, because I don't think it could be possible to detect stars more distant than 5000 pc, or am I wrong in this supposition ?

And I've rejected all Tycho stars only with a new calculated parallax from the B-V color index < 0.0002 " for the same reason...

Do you think it is a mistake ?

Pascal

Pascal,

It seems to me that it is reasonable to omit stars which have systematic errors that are large compared to their parallax values. Certainly the errors are larger for more distant stars as the parallax gets smaller and more difficult to measure.

What is meant by "large" is a matter of opinion and I think you should make that decision. For example, 15% seems large to me, and 30% feels unreasonable.

I hope this helps a little.

Rigel wrote:in V1.2 I've put Hipparcos stars with a parallax < 0.0002" to this limit value, because I don't think it could be possible to detect stars more distant than 5000 pc, or am I wrong in this supposition ?

No, you're right! But that doesn't mean you must exclude this star from the catalogue, for reasons I explained in the previous post. It may not be possible to detect stars 5000 pc away in Tycho, but this star was detected! This means that the star is not 5000 pc away, but much closer. I'll make a try to put the logic simple:

A) The Tycho catalogue is magnitude limited, not distance limited. It is only complete to about V=10.5.

B) A class III source of V=10.5 and (for example!!) B=12.0 B-V=1.5 absolute V magnitude ~ -1.5, this translates to a distance of about 2000 pc. Thus, class III sources of B-V=1.5 can be seen out to a distance of about 2000 pc in the Tycho catalogue.

C) A class V source of V=10.5 and (for example!!) B=12.0 B-V=1.5 absolute V magnitude ~ 9, this translates to a distance of about 20 pc. Thus, class V sources of B-V=1.5 can be seen out to a distance of about 20 pc in the Tycho catalogue.

Assume, for the sake of argument, that you find a star at the limiting magnitude V=10.5 with B-V=1.5 and the oddly small parallax 0.0002" with standard error 0.010". Should you disregard this star? No! Because even if you add the parallax error to the parallax, and get the parallax 0.0102", the corresponding distance is still 100 pc. That is, we say that the star with one-sigma confidence is further away than 100 pc. Similarly, we have that on the two-sigma level (parallax 0.0202"), the star is further away than 50 pc. But according to C) we wouldn't be able to see this star if it was of class V and at the distance of 100 pc or even 50 pc, because the furthest we can see class V sources with B-V=1.5 is 20 pc. Thus the star is not of class V but class III => we derive with B) the "colour" distance 2000 pc corresponding parallax 0.0005" which is consistent with 0.0002"+/-0.010".

I hope you now see that it is generally wrong to dismiss stars with small parallaxes. We only use the parallaxes to differentiate between class III and class V stars, and then a small parallax tells us at least that the star is distant, which is useful information.

Rigel wrote:And I've rejected all Tycho stars only with a new calculated parallax from the B-V color index < 0.0002 " for the same reason...

Well... The faintest stars in Tycho have about V=12. A parallax of 0.0002" a distance of 5000 pc an absolute magnitude of Mv = -1.5. A smaller parallax an even brighter absolute magnitude (Mv < -1.5), but I don't see how you could possibly deduce such a bright absolute magnitude. According to the fit P3 you sent me, the brightest Mv seems to be ~ -1. To deduce a larger distance than 5000 pc from a V = 12 mag star you need the absolute magnitude to be Mv < -1.5. If you do find stars with deduced "colour" parallaxes as small or smaller than 0.0002", this suggests an error somewhere in your code!

Regarding finding a fitting function Mv = f(B-V) for various classes, I think the best you can do is to use the Hipparcos data (as you've already done). There will always be a spread due to, for example, variable extinction, which you cannot correct for with only B-V colours. That's why it's important to find the RMS of your fit to determine how accurate your "colour" parallaxes are. With more colours it would be possible to correct for extinction as well and get a more accurate fit, but this involves more work since you need to check other sources than the Hipparcos and Tycho catalogues and make (somewhat difficult) cross identifications. Currently I estimate that the derived distances using your procedure have an error of about 50%. Not too bad

OK, I realise that all this may seem like difficult stuff, and maybe I'm not too good at explaining it, but don't give up. You've come a long way already, so don't hesitate to ask me about specifics. And the reward is a larger and more accurately simulated universe!

/Alexis

chris wrote:

dtessman wrote:Chris,

is the stars.dat format the same in the next version?

Yes, it's the same format as before . . . any suggestions on improving it? I do plan on keeping the primary star catalog file format binary, both for compactness and to keep loading fast.

--Chris

The addition of other star types would be nice. I am no astronomer, so this may not make any sense, but it would be nice to use something like light years for distance, rather than parallax. Would that get rid of parallax error as well?

Definitely keep the stars.dat in binary! XML should be used for import, export, and add-ons only. I asked, so that I could have a new version of CStarsConv ready.

Wow, fast server you got there. I got a nice 150 k/sec.