ASCC

[t00fri]

I have 'ThinkerX' distances for something on the order of a thousand plus stars that also have Hip parallaxes accurate to within 5%.

That's a great sample for a reliability/error study.

[Aside: I hope you have some powerful plot programs and use e.g. Perl to do extensive calculations on MANY data. Excellent math frameworks are also Maple / Mathematica and Matlab, but they cost a lot of money. I have them from my laboratory...Perl is free and most powerful. Most database calculations for Celestia we do with Perl]

The first thing I would do with your sample is to plot your ThinkerX distances (x-axis) versus the corresponding accurate HIP distances (y-axis). That correlation plot tells you already a lot about inherent systematics. Ideally you should see data points that are scattered regularly around a straight line

HIP distance = ThinkerX distance ..............................................(1)

Any shifts or curvatures visible in a best fit to the correlated data indicate systematic errors.
As a next step you could add estimated error bars in x and y (5% for HIP), which turns the data points into crosses. Inspecting the new plot as before, you can estimate the significance of possible deviations from the straight line (1).

Then you could go on (depending on your technical abilities...) to do a least-square fit -- with a professional-level fitting program -- of the correlation data to the best straight line. The program will then generate numerical errors for you....

Good luck and much fun

Fridger

[starguy84]

I have a suspicion this bears little resemblance to any other photo spectroscopic distance scheme you have seen. However, it did well enough in the callibrations - and so far its matched up reasonably well with what few alternative distances I've found for the stars in the ASCC.

I admit it isn't... I still haven't decided if you're effectively doing linear interpolation between spectral types with your corrections, have found a way to correct for metallicity/luminosity differences, or... other. Right now, my money's on linear interpolation: Your formula,

Code: Select all

Absolute V Mag = 5.1-(5*0.68)+(5*BmV)    which is effectively
Absolute V Mag = 5.1 + 5*(BmV-0.68)

suggests that a star with a higher B-V (ie, redder) than normal should have a larger (ie, fainter) absolute magnitude than normal. The main sequence is not linear, which would explain why your relations break down for large differences between the dead-average B-V and the measured B-V.

Aside: I hope you have some powerful plot programs and use e.g. Perl to do extensive calculations on MANY data. Excellent math frameworks are also Maple / Mathematica and Matlab, but they cost a lot of money.

I suppose I could subtract my distances from the Hip distances, convert the result to absolute numbers (the statistical error), and then tally them up (ye Gods thats a lot of arithmetic!) and average them out for some sort of overall statistical error.

To be honest, Excel (or OpenOffice Calc, which is free and vaguely similar) should be sufficient to make plots and linear fits, and might be easier to use in the short term. While I can't say much about the scientific robustness of their tools (I prefer ITT IDL), Excel has SLOPE, INTERCEPT, CORREL, and RSQ functions http://phoenix.phys.clemson.edu/tutorials/excel/stats.html, OpenOffice Calc has a LINEST function that does a linear regression, as well as SLOPE, INTERCEPT, CORREL and RSQ functions similar to Excel's.
http://wiki.services.openoffice.org/wiki/Documentation/How_Tos/Calc:_LINEST_function

[ThinkerX]

Been busy trying to sort out the double star stuff with my distance scheme. Thus far, it looks like the wide multiple stars are not that much of a problem...but the close doubles...those can muck things up...sometimes.

[Aside: I hope you have some powerful plot programs and use e.g. Perl to do extensive calculations on MANY data. Excellent math frameworks are also Maple / Mathematica and Matlab, but they cost a lot of money. I have them from my laboratory...Perl is free and most powerful. Most database calculations for Celestia we do with Perl]

Where do I find this 'Perl' at?

admit it isn't... I still haven't decided if you're effectively doing linear interpolation between spectral types with your corrections, have found a way to correct for metallicity/luminosity differences, or... other. Right now, my money's on linear interpolation:

Actually, my money is on linear interpolation as well.

Your formula,

Code: Select all
Absolute V Mag = 5.1-(5*0.68)+(5*BmV) which is effectively
Absolute V Mag = 5.1 + 5*(BmV-0.68)

suggests that a star with a higher B-V (ie, redder) than normal should have a larger (ie, fainter) absolute magnitude than normal. The main sequence is not linear, which would explain why your relations break down for large differences between the dead-average B-V and the measured B-V.

That is pretty much the idea. I started with the basic formulas for determining distance moduli and determining distance from that. The examples I ran across all assumed that a star of a given spectral type would have a single fixed number - say 4.7 for G2V stars. But...absolute magnitudes can vary pretty radically within a given spectral type, by 0.5 or better. Hence, even a 'normal' G2V star could have an absolute magnitude anywhere from 4.2 to 5.2. So...if you applied the normal dead average of 4.7 to a star that really had an absolute magnitude of 4.2...well, then your distance would be off by 25%. If the stars real absolute magnitude was say...4.5, you'd only be off by 10%.

So...I started looking for a way to tweak the average absolute magnitude in the right direction. I noticed right off that the small B-V numbers tended to be associated with stars of the 'higher' spectral types. I also noticed this applied to V-J and J-H as well, so I started wondering if it would apply to stars within the same spectral type as well. It does...though it is more of a tendency than a hard and fast trend. At times I regard it as 'loading the dice' - the B-V, V-J, and J-H numbers will tend to move in the right direction, but not always. But if say B-V stays put or even moves the 'wrong' way, the 'loaded dice' effect is still in place for the other two, which is often all it takes.

Normally, when putting the callibration tables together, I would sort the stars by their true (Hip distance determined) absolute magnitude, with my hits and misses marked in special fields. Sorted this way, almost all the misses would be towards the top (the bright super luminals) and the very bottom (subluminous stars), with just a very few other misses in between. When I would sort by one of my derived absolute magnitudes though, that picture would change pretty radically: the misses and hits would all be jumbled together. While the V-J and J-H filters would work to weed out the giants, I never could spot more than a handful - say 25% or so - of the superluminious and subluminous stars that caused such problems.

However and whatever, the bottom line is about 75% of the time it will come up with a distance accurate to within 20%, and about 60% of the time it will cough up a distance accurate to within 10%. That seems to be reasonably good, roughly on a par with Weis, though maybe not as good as the current top notch photometric methods.

To be honest, Excel (or OpenOffice Calc, which is free and vaguely similar) should be sufficient to make plots and linear fits, and might be easier to use in the short term. While I can't say much about the scientific robustness of their tools (I prefer ITT IDL), Excel has SLOPE, INTERCEPT, CORREL, and RSQ functions http://phoenix.phys.clemson.edu/tutoria ... stats.html, OpenOffice Calc has a LINEST function that does a linear regression, as well as SLOPE, INTERCEPT, CORREL and RSQ functions similar to Excel's.

I will have to look into this.

[t00fri]

Perl is part of any UNIX/Linux system and thus of MAC OSX, too. It also exists for Windows (Active Perl, Stawbery Perl,...) . It's free, of course. Installation of additional modules from the CPAN world archive is done with a few clicks. There is no better scripting language for handling MANY data or any textual stuff. There are many excellent tutorials out there... Perl is easy to learn, despite its enormous power.

E.g. a > 1000 line Perl code processes my 10000+ galaxies in about 15 seconds, reading out ~ 10 catalogs (!), merging them, doing some heavy transformations from the skyplane to Celestia's excliptic plane and reducing the extended deVauxcouleurs morpholgy classes to the standard Hubble morphology, etc.

Fridger

[starguy84]

But if say B-V stays put or even moves the 'wrong' way, the 'loaded dice' effect is still in place for the other two, which is often all it takes.

Some caution needs to be taken with this: it is possible (but not 100% reliable) to distinguish between giants and dwarfs using photometric colors.

Basically, the way we can distinguish between giants and dwarfs using spectroscopy is that stars of different luminosities and surface gravities have slightly but detectably different spectral absorption lines. Photometric colors are a record of the light coming out of the star in a particular wavelength range* so a star with different spectral lines (ie, giant) can have different colors than a dwarf. You can't see that in one color, but if you plot color versus color, occasionally you'll get a noticeable spread. Dwarfs of a particular B-V color should have similar V-J colors; if they DON'T that could be a sign that they're giants. Exactly what "difference" is meaningful, is something you'd have to figure out. I happen to know from my own thesis work that V-K versus J-K discriminates M0-M7 pretty well.

Then again, for MOST stars, a reduced proper motion diagram will split out your superluminous stars for you, and then you can make separate relations for giants assuming they're all luminosity class III (Hipparcos doesn't have good distances to many/any class I supergiants anyway). The distances will be a lot less accurate, judging from the scatter I've seen in the giant branch, but if ASCC has as many giants as you suggest it might be worth doing.

Actually, my money is on linear interpolation as well.

That might be grounds for doing actual linear interpolation across your tables... cut out the spectral type and just interpolate between your B-V=MV (and so on) anchors.

Been busy trying to sort out the double star stuff with my distance scheme. Thus far, it looks like the wide multiple stars are not that much of a problem...but the close doubles...those can muck things up...sometimes.

2MASS used 2" pixels (http://vizier.cfa.harvard.edu/viz-bin/Cat?II/246) at sites where the seeing is better than that, so 2" is the limit on 2MASS photometry. Tycho-2 claims to include double star components as close as 0.8" http://vizier.cfa.harvard.edu/viz-bin/Cat?I/259. It's therefore possible to have two objects with separate B and V, but combined JHK. These can (in theory) be split using B-V colors, but you would have to assume standard dwarf JHK colors. You're liable to get wrong colors from giants and places where there's lots of dust, which would be the galactic plane... and that's where you are most likely to have close doubles, unfortunately.

for B1-V1 find V-J and V-H and V-K
with V1, solve for J1, H1, and K1
for B2-V2 find V-J and V-H and V-K
with V2, solve for J2, H2, and K2
check:

Code: Select all

H + 2.5 log(10^(-H1/2.5)+10^(-H2/2.5)) = (small photometric scatter)

etc.
If those equations do NOT work, one or both of the stars do not have normal JHK colors, and you should throw both out.

E.g. a > 1000 line Perl code processes my 10000+ galaxies in about 15 seconds, reading out ~ 10 catalogs (!), merging them, doing some heavy transformations from the skyplane to Celestia's excliptic plane and reducing the extended deVauxcouleurs morpholgy classes to the standard Hubble morphology, etc.

To be honest, I'm coming around to Fridger's side of the argument regarding the statistics plots... Neither OpenOffice.org calc nor Excel are really set up for this. For instance, I have the most recent version of OpenOffice.Org Calc and it only allows 65536 rows; you will NOT be able to handle all of ASCC at once. If you load everything into Calc, you'll have to break it up into at least 39 pieces. I found a TREND() that MIGHT be a decent linear interpolator [url][http://www.openofficetips.com/blog/archives/2005/10/basic_functions_15.html/url] but I'm not sure. And I don't know much about Excel, but I doubt it's any better. [url]http://www.blueleafsoftware.com/Products/Dagra/LinearInterpolationExcel.php[/url]

*minus absorption/scattering from the interstellar medium, Earth's atmosphere, the actual filter, and whatever's recording the photons, but bear with me

[t00fri]

But if say B-V stays put or even moves the 'wrong' way, the 'loaded dice' effect is still in place for the other two, which is often all it takes.

Some caution needs to be taken with this: it is possible (but not 100% reliable) to distinguish between giants and dwarfs using photometric colors.

Basically, the way we can distinguish between giants and dwarfs using spectroscopy is that stars of different luminosities and surface gravities have slightly but detectably different spectral absorption lines. Photometric colors are a record of the light coming out of the star in a particular wavelength range* so a star with different spectral lines (ie, giant) can have different colors than a dwarf. You can't see that in one color, but if you plot color versus color, occasionally you'll get a noticeable spread. Dwarfs of a particular B-V color should have similar V-J colors; if they DON'T that could be a sign that they're giants. Exactly what "difference" is meaningful, is something you'd have to figure out. I happen to know from my own thesis work that V-K versus J-K discriminates M0-M7 pretty well.

Then again, for MOST stars, a reduced proper motion diagram will split out your superluminous stars for you, and then you can make separate relations for giants assuming they're all luminosity class III (Hipparcos doesn't have good distances to many/any class I supergiants anyway). The distances will be a lot less accurate, judging from the scatter I've seen in the giant branch, but if ASCC has as many giants as you suggest it might be worth doing.

Actually, my money is on linear interpolation as well.

That might be grounds for doing actual linear interpolation across your tables... cut out the spectral type and just interpolate between your B-V=MV (and so on) anchors.

Been busy trying to sort out the double star stuff with my distance scheme. Thus far, it looks like the wide multiple stars are not that much of a problem...but the close doubles...those can muck things up...sometimes.

2MASS used 2" pixels (http://vizier.cfa.harvard.edu/viz-bin/Cat?II/246) at sites where the seeing is better than that, so 2" is the limit on 2MASS photometry. Tycho-2 claims to include double star components as close as 0.8" http://vizier.cfa.harvard.edu/viz-bin/Cat?I/259. It's therefore possible to have two objects with separate B and V, but combined JHK. These can (in theory) be split using B-V colors, but you would have to assume standard dwarf JHK colors. You're liable to get wrong colors from giants and places where there's lots of dust, which would be the galactic plane... and that's where you are most likely to have close doubles, unfortunately.

for B1-V1 find V-J and V-H and V-K
with V1, solve for J1, H1, and K1
for B2-V2 find V-J and V-H and V-K
with V2, solve for J2, H2, and K2
check:

Code: Select all

H + 2.5 log(10^(-H1/2.5)+10^(-H2/2.5)) = (small photometric scatter)

etc.
If those equations do NOT work, one or both of the stars do not have normal JHK colors, and you should throw both out.

E.g. a > 1000 line Perl code processes my 10000+ galaxies in about 15 seconds, reading out ~ 10 catalogs (!), merging them, doing some heavy transformations from the skyplane to Celestia's ecliptic plane and reducing the extended deVauxcouleurs morpholgy classes to the standard Hubble morphology, etc.

To be honest, I'm coming around to Fridger's side of the argument regarding the statistics plots... Neither OpenOffice.org calc nor Excel are really set up for this. For instance, I have the most recent version of OpenOffice.Org Calc and it only allows 65536 rows; you will NOT be able to handle all of ASCC at once. If you load everything into Calc, you'll have to break it up into at least 39 pieces. I found a TREND() that MIGHT be a decent linear interpolator [url][http://www.openofficetips.com/blog/archives/2005/10/basic_functions_15.html/url] but I'm not sure. And I don't know much about Excel, but I doubt it's any better. [url]http://www.blueleafsoftware.com/Products/Dagra/LinearInterpolationExcel.php[/url]

*minus absorption/scattering from the interstellar medium, Earth's atmosphere, the actual filter, and whatever's recording the photons, but bear with me

[ThinkerX]

Basically, the way we can distinguish between giants and dwarfs using spectroscopy is that stars of different luminosities and surface gravities have slightly but detectably different spectral absorption lines. Photometric colors are a record of the light coming out of the star in a particular wavelength range* so a star with different spectral lines (ie, giant) can have different colors than a dwarf. You can't see that in one color, but if you plot color versus color, occasionally you'll get a noticeable spread. Dwarfs of a particular B-V color should have similar V-J colors; if they DON'T that could be a sign that they're giants. Exactly what "difference" is meaningful, is something you'd have to figure out. I happen to know from my own thesis work that V-K versus J-K discriminates M0-M7 pretty well.

I spent quite a bit of time comparing B-V, V-J, and J-H interrelationships trying to spot giants. Usually, it was along the lines of looking for stars where V-J or J-H were unusually low or high, either compared against the norm or against each other. This did let me spot some superlumious stars...some of the time, but it also yielded so many false positives I eventually abandoned it. The 'occasionally' about sums it up.

Then again, for MOST stars, a reduced proper motion diagram will split out your superluminous stars for you, and then you can make separate relations for giants assuming they're all luminosity class III (Hipparcos doesn't have good distances to many/any class I supergiants anyway). The distances will be a lot less accurate, judging from the scatter I've seen in the giant branch, but if ASCC has as many giants as you suggest it might be worth doing.

Pretty much what I'm aiming at, combined with the B-V, V-J, and J-H cuts. Some will still sneak in (mostly class IV stars), but the ones that do will mimic everything else so closely that my system will actually provide decent distances to them about half the time.

2MASS used 2" pixels (http://vizier.cfa.harvard.edu/viz-bin/Cat?II/246) at sites where the seeing is better than that, so 2" is the limit on 2MASS photometry. Tycho-2 claims to include double star components as close as 0.8" http://vizier.cfa.harvard.edu/viz-bin/Cat?I/259. It's therefore possible to have two objects with separate B and V, but combined JHK. These can (in theory) be split using B-V colors, but you would have to assume standard dwarf JHK colors. You're liable to get wrong colors from giants and places where there's lots of dust, which would be the galactic plane... and that's where you are most likely to have close doubles, unfortunately.

If I understand correctly, this explains something else I noticed in the callibrations: with some (but not all) of the close doubles, the distance moduli for V would be *much* closer than the combined distance moduli (the J and H distance moduli would be so far out, the V distance moduli couldn't do much to correct them). I actually considered using just the V distance moduli for some close doubles, instead of the combined moduli, but ultimately decided that A) I didn't know enough about close doubles; and B) this happened only sometimes; so I decided to go with letting the combined distance moduli stand and with flagging close binaries as possibly having suspect distances.

To be honest, I'm coming around to Fridger's side of the argument regarding the statistics plots... Neither OpenOffice.org calc nor Excel are really set up for this. For instance, I have the most recent version of OpenOffice.Org Calc and it only allows 65536 rows; you will NOT be able to handle all of ASCC at once. If you load everything into Calc, you'll have to break it up into at least 39 pieces. I found a TREND() that MIGHT be a decent linear interpolator [url][http://www.openofficetips.com/blog/archives/2005/10/basic_functions_15.html/url] but I'm not sure. And I don't know much about Excel, but I doubt it's any better. [url]http://www.blueleafsoftware.com/Products/Dagra/LinearInterpolationExcel.php[/url]

More goodies to track down and play with.

I am actually downloading partly presorted chunks of the ASCC in units of 9999; and anticipate something on the order of 15 - 20 such sections total. I expect to get around 40,000 - 50,000 star distances out of all that. (Currently, I've downloaded and processed four 'blocks'). I expect to finish up around late March or early April.

As to what I'm using...would you believe a somewhat elderly (?) MS Database Works?

::watches Starguy run screaming off into the night::

[starguy84]

As to what I'm using...would you believe a somewhat elderly (?) MS Database Works?

::watches Starguy run screaming off into the night::

AAAAAaaaaaaaah

Seriously, a relatively simple script will be able to process everything you need very quickly, seeing as you only need to do the same few operations for each star. You set up the script, you pipe all your data through it, you're done.

Code: Select all

For a single set of name, ra, dec, PMra, PMdec, B, V, J, H(, K?)
Compute B-V, V-J, J-H
Compute Reduced Proper motion H
Determine luminosity class, switch to correct distance lookup table
Calculate MV based on B-V lookup, MJ based on V-J lookup, MH based on J-H lookup
Look up spectral type
Calculate distance moduli
(if systematic errors are already available) calculate Distance and distance error
(if distance error is small enough) print name, ra, dec, V, spectral type+luminosity class, distance, (distance error) to a file in a format Celestia can read
repeat 2.5 million times

As an aside, what VizieR table are you using? It's not Kenyon & Hartmann 1995 (J/ApJS/101/117/tablea5); they don't actually list the MV for dwarf stars of a particular spectral type and they would have predated Hipparcos even if they had... I'm not aware of any others.

If it includes V-K or B-K colors, I'd suggest you take a look at them. B and K are very different colors; they make a very long lever arm to measure the changing peaks of spectra. V-K is also good too, although maybe not so much for blue stars.

[ThinkerX]

Seriously, a relatively simple script will be able to process everything you need very quickly, seeing as you only need to do the same few operations for each star. You set up the script, you pipe all your data through it, you're done.

I have thought about something like that. But...

A) My programming skills are somewhere between feeble and nonexistent; and

B) The result is only as good as the sum of the data entered. Or to put it another way, while the approach you describe is fast, it is also fast enough to where things can and do get missed.

I suspect the Ukrainians that put the ASCC together did do something like that, or maybe used a program like the one tOOfri described. And it shows: the ASCC is missing about 75% of the HD numbers it should have, along with a small but signifcant percentage of DM numbers. The double star flags have...problems...and fixing/providing double star info is something that has occupied much of my time these last couple of weeks. (On the bright side, I did manage to collect enough measures for some of these stars to where somebody with the time and the know-how - or possibly just the right computer program - could work up some rudimentary orbits for some of these systems).

As an aside, what VizieR table are you using? It's not Kenyon & Hartmann 1995 (J/ApJS/101/117/tablea5); they don't actually list the MV for dwarf stars of a particular spectral type and they would have predated Hipparcos even if they had... I'm not aware of any others.

Don't quite know how to link directly to the tables in VizeR. However, the authors name is Tsvetkov, and the table is one of two in 'New Spectral Types for Tycho Stars'. Its the first of these tables, titled `Tabulated B-V color indices and absolute magnitudes M...' I usually get to it by running a search for 'Tycho'; its one of the last on the list.

Anyhow, the table gives dead average B-V and absolute V magnitudes for stars of all spectral types from O5 to M5, and of luminosity classes I, III, and V. The author claims to have done a B-V and spectral type V averaging scheme on something like a third of the Hip catalogue to compile this chart, but even so, I find myself wondering just how he managed to assemble enough accurate data to put it together (for example...how many high accuracy Hip parallaxes are there for luminosity class I stars).

I suppose one of these days...maybe next year, maybe sooner, maybe later, I'll break down and do some templates for M stars. Maybe use those templates on the NLTT, since it purportedly has a very large number of M stars; might get several thousand distances out of it. Would your crew be interested? Probably about on a par with Weis, accuracy wise.


In the note section for the catalogue (or was it the abstract of the paper?) the author talked about something directly relevant here: a spectroscopic distance catalogue he made of some 30,000+ Tycho/ASCC stars of luminosity classes III and V...but that catalogue did not make it into Vizier.

[John Van Vliet]

--- edit ---

[starguy84]

The result is only as good as the sum of the data entered. Or to put it another way, while the approach you describe is fast, it is also fast enough to where things can and do get missed.

Fair enough, but there's also something to be said for the consistency of an automated solution.
The Tycho-2 spectral type database doesn't mention the HD catalog. Odd. So, they must be coming into ASCC third-hand. Maybe that's why they were missing.

Garbage in, Garbage out, as you said.

Don't quite know how to link directly to the tables in VizeR. However, the authors name is Tsvetkov, and the table is one of two in 'New Spectral Types for Tycho Stars'.

Surprisingly, it's not in the NASA Astrophysics Data System... but Google found the paper. I have institutional access so I can read it, and the seminal bits are:
1.) Their method is "taken directly from Mihalas & Binney 1981" (Galactic Astronomy: Structure & Kinematics, second edition: a seminal textbook on the subject) with a table supposedly taken from Allen & Cox 2000 (Allen's Astrophysical Quantities, a standard reference). The method is surprisingly similar to yours, although I couldn't find any formuale in Mihalas & Binney 1981:

Code: Select all

E(B-V) = (B-V)_observed - (B-V)_tabulated
R= 3.30 + 0.28*(B-V)_tabulated + 0.04*E(B-V)
A_v = R * E(B-V)
log(distance) = 0.2 (V_observed - MV_tabulated - A_v + 5)


2.) They apparently assume stars come in a few discrete flavors with all variation due to reddening (even NEGATIVE reddening, which they agree is impossible, and ignore)
3.) They have used corrections from Tycho-V and Tycho-B to Johnson-V and Johnson-B that "are believed to be not accurate enough", quoting a collaborator of theirs who knows better... I've also seen other calibrations from, for instance, Bessel 2000 PASP (a colleague of Weis's, whom I have met... His work is in textbooks and our 'standard' calibrations, but he was glad to meet ME, because apparently nobody my age knows who he is)
4.) They find that their methods systematically UNDERestimate distances to main sequence stars, and OVERestimate distances to giants, which they suspect is due to systematic errors in the Hipparcos parallaxes.
5.) They do not seem to mention where this massive catalog of 35,000 stars ARE. I can't find either Popov & Tvetskov 2006 or Smirnov 2006, where that spectrophotometric distance catalog might be.

Quite frankly, I prefer your method (if nothing else, because you're actually going to produce a catalog everyone can enjoy). Interstellar reddening IS a problem, but it increases with DISTANCE and is usually negligible nearby... probably out to 100 pc it can be ignored, but don't quote me on that.

As a practical concern for you, reddening is accompanied by ~3.1 times as much dimming (at least, that's what I learned in grad school, not 3.3... and yes, it does depend on where and what the large dust clouds actually are in space), so it will make the stars systematically redder and (much) fainter as you go farther out. Your photometric distances will be larger than the actual distance if the slope of the main sequence is < 3.1 (ie, getting redder causes LESS change in absolute magnitude); they'll be less if it's > 3.1 (ie, getting redder drops the absolute magnitude more than 3.1x). This should show up in the ThinkerX vs Hipparcos plots that Fridger asked for.

I suppose one of these days...maybe next year, maybe sooner, maybe later, I'll break down and do some templates for M stars. Would your crew be interested? Probably about on a par with Weis, accuracy wise.

I'd be more interested in those objects possibly within 50 pc...

[ThinkerX]

1.) Their method is "taken directly from Mihalas & Binney 1981" (Galactic Astronomy: Structure & Kinematics, second edition: a seminal textbook on the subject) with a table supposedly taken from Allen & Cox 2000 (Allen's Astrophysical Quantities, a standard reference). The method is surprisingly similar to yours, although I couldn't find any formuale in Mihalas & Binney 1981:

Code: Select all
E(B-V) = (B-V)_observed - (B-V)_tabulated
R= 3.30 + 0.28*(B-V)_tabulated + 0.04*E(B-V)
A_v = R * E(B-V)
log(distance) = 0.2 (V_observed - MV_tabulated - A_v + 5)

...'taken directly from Mihalas and Binney 1981'...well, two of the formulas you cited there do look like the ones I dug up (from Wiki, no less) when first starting this project. The 'R=3.30 +0.28*(B-V)_tabulated +0.04*E(B-V)' looks like a correction for interstellar reddening, which was something I was interested in but couldn't find, at least in a understandable format.

2.) They apparently assume stars come in a few discrete flavors with all variation due to reddening (even NEGATIVE reddening, which they agree is impossible, and ignore)

So...if I understand this right, they assume the 'dead average absolute magnitudes' in their chart are so close to the actual absolute magnitudes that no further correction is necessary. I see serious problems here.

3.) They have used corrections from Tycho-V and Tycho-B to Johnson-V and Johnson-B that "are believed to be not accurate enough", quoting a collaborator of theirs who knows better... I've also seen other calibrations from, for instance, Bessel 2000 PASP (a colleague of Weis's, whom I have met... His work is in textbooks and our 'standard' calibrations, but he was glad to meet ME, because apparently nobody my age knows who he is)

I had doubts about this as well...which is why I used only those stars with very low margins of error on the magnitudes. 'Scatter' did and still does bother me, but I guess I'll just have to live with that.

You meant Bessel, eh? I suspect I might be closer to Bessels age than yours - my college days ended a good two dozen years ago. I just do projects like this now and again to pass the time during the long cold dark winters we have here.

4.) They find that their methods systematically UNDERestimate distances to main sequence stars, and OVERestimate distances to giants, which they suspect is due to systematic errors in the Hipparcos parallaxes.

I didn't do any thing with giant stars, except try to exclude them, so I can't say anything there...but from what I can tell, my system does not systematically underestimate the distances to normal main sequence stars...except where the superluminals are concerned. For those stars...about 5-15% of the total, depending on the spectral type, my distances do fall short, sometimes by way over 50%. I wonder if my ...linear?... system actually fixes some of the problems with theirs?

The second table they had in Vizier dealt with correcting what they saw as Hip errors and adjusting spectral types. They seemed to imply there was only a relatively small number of true problem cases...but I've wondered about that.

5.) They do not seem to mention where this massive catalog of 35,000 stars ARE. I can't find either Popov & Tvetskov 2006 or Smirnov 2006, where that spectrophotometric distance catalog might be.

I'd like to know as well. I suspect it may have been exluded from VizieR with good reason, though.

Quite frankly, I prefer your method (if nothing else, because you're actually going to produce a catalog everyone can enjoy).

Thank You!

Interstellar reddening IS a problem, but it increases with DISTANCE and is usually negligible nearby... probably out to 100 pc it can be ignored, but don't quote me on that.

As a practical concern for you, reddening is accompanied by ~3.1 times as much dimming (at least, that's what I learned in grad school, not 3.3... and yes, it does depend on where and what the large dust clouds actually are in space), so it will make the stars systematically redder and (much) fainter as you go farther out. Your photometric distances will be larger than the actual distance if the slope of the main sequence is < 3.1 (ie, getting redder causes LESS change in absolute magnitude); they'll be less if it's > 3.1 (ie, getting redder drops the absolute magnitude more than 3.1x). This should show up in the ThinkerX vs Hipparcos plots that Fridger asked for

I was aware interstellar reddening could become a serious problem at larger distances (I've also seen comments in various papers that it doesn't become a significant factor until after 100 parsecs).

Thus far, the rough and ready break down works out something like this:

The K stars are almost never past 100 PC. Actually, only some of the K0 and K2 stars have distances past 50 parsecs...though I'm starting to wonder about a few of the supposedly nearer ones
because they are close binaries.

The G stars range from 50 - 180 parsecs, give or take. Very few much closer or further out. If it stopped here...I could almost ignore interstellar reddening altogether. But...

The F stars tend to have distances from around 60 parsecs out to about 220 parsecs. Some of the brighter ones (F5's) turn up with distances past 250 parsecs...but that is almost the limit.

The 'target' effect does bother me somewhat...enough to the point where I've actually considered imposing a hard 150 or 180 parsec limit or so on the catalogue. If I did that, I *might* be able to ignore reddening altogether...but I'd also be chucking about a quarter of the stars in the catalogue. Gotta give that one a lot more thought.

I'd be more interested in those objects possibly within 50 pc...

Also my primary interest. Years ago, I downloaded something called the HYG (Hip/Yale/Gliese)catalogue (this was before discovering VIzieR) and spent a lot of time adding to it and playing around with it before realizing that catalogue was too badly flawed to be of much real use (no error bars on the distances and no proper motions).

However...in mucking around with it, I came across what I thought of as the 'mystery of the missing stars'. Basically, I tallied up the stars within 5 parsecs and did a volume calculation, assuming that the stars within that distance were all known, with good distances. I then used this to 'measure' star density out past this distance. The numbers started getting screwy past about 10 parsecs...and at 25 parsecs...well, it looked like the HYG, at least, was missing something on the order of 50% (3500 or so) of the stars it should have had. When I ran the calculation out to 50 parsecs (pretty much the limit of the HYG) the numbers became even more lopsided; it looked like something on the order of 80% (around 40,000, if memory serves) of the stars were missing. This was when I started tracking down photometric distance catalogues and eventually blundered across Weis's work.

Now...granted the vast majority of these 'missing' stars are probably M class dwarfs or fainter...but when I started comparing spectral types (emphasis on at least vaguely sun like stars - F5 to K8, absolute magnitudes 3.5 to 8.5) - it looked like a lot of those were missing from the catalogues as well past 10 parsecs - probably at least several hundred of them. So I gave it a bit of thought. The Hip catalogue nailed pretty much everything with an apparent magnitude of 7.5 or greater. Given their absolute magnitudes, that pretty much ruled out the odds of the 'missing' sun like stars being F's, or even brighter G's. That left the fainter G's and the K's. What very few active parallax programs there are these days concentrate almost exclusively on very faint stars, and the data from Gaia probably won't be available for a decade yet. Hence...one of the reasons for this project.

[starguy84]

I wonder if my ...linear?... system actually fixes some of the problems with theirs?

Their system is based on three assumptions I don't agree with: 1.) That all G2V stars are absolutely identical, 2.) That any differences are due to reddening, 3.) they don't allow any negative corrections. Your linear corrections to colors are based on fitting the data around each spectral type/color point, and do not make those assumptions. I think your system is better.

Of course, the icing on the cake is that they went on to suggest the disagreements with Hipparcos trig parallaxes were a problem with Hipparcos... Hipparcos surely isn't perfect, but they offered no other proof or evidence to support that claim.

I suspect it may have been exluded from VizieR with good reason, though.

I think there's a bias in NASA ADS and VizieR toward countries that publish in English- North America, South America, ESA member nations... The paper you found was translated from Russian and republished in another European journal; it's possible the others were simply never translated or republished.

The 'target' effect does bother me somewhat...enough to the point where I've actually considered imposing a hard 150 or 180 parsec limit or so on the catalogue. If I did that, I *might* be able to ignore reddening altogether...but I'd also be chucking about a quarter of the stars in the catalogue. Gotta give that one a lot more thought.

Distance-dependent reddening correction? (once you know you're largely free of other systematic effects) I don't know what form that would specifically take...

However...in mucking around with it, I came across what I thought of as the 'mystery of the missing stars'. Basically, I tallied up the stars within 5 parsecs and did a volume calculation, assuming that the stars within that distance were all known, with good distances. I then used this to 'measure' star density out past this distance. The numbers started getting screwy past about 10 parsecs...and at 25 parsecs...well, it looked like the HYG, at least, was missing something on the order of 50% (3500 or so) of the stars it should have had. When I ran the calculation out to 50 parsecs (pretty much the limit of the HYG) the numbers became even more lopsided; it looked like something on the order of 80% (around 40,000, if memory serves) of the stars were missing.

Ah, the missing stars. Your analysis is exactly right, as I understand the problem.

[ThinkerX]

Ok..I gave Tsvetkov's chart a bit more thought, which in turn prompted me to go take another look at my callibration tables, which resulted in a bit more thought. I believe I see what he was wrestling with.

First off, his system is based pretty much exclusively on V. The 'reddening' modifier...it looks like it serves a very roughly equivilent function to my multiplier, which means my distance moduli for V and his should be pretty close.

That gets him into problems, though, because while that yields fairly accurate distances for...call it 0.4 either way up and down from the dead average magnitude, it falls off pretty dramatically past that - especially going 'up'. In my callibrations, the V magnitude tends towards the fainter (less bright) end of the scale. But there are an awful lot of stars brighter than 0.4 above the dead average.

Now...at this point I have the distance moduli for J and H magnitudes to step in. The J distance moduli gives accurate but spotty distances from 0.6 towards the faint end of the scale and as much as 0.8 or so towards the bright side. Likewise, the H distance moduli, overall, tends to run a bit towards the hot side of the scale. If you had to pick one of the three for a star you suspected to be a bit brighter than usual, then either J or H would be a better pick than V for a more accurate distance. Averaged together, my distances are...well...not skewed exactly...but give accurate distances further 'up' than they do going 'down'. OR...I've said before that my system seems to do reasonably good 0.5 either way, and with only spotty hits past that - but almost all of those spotty hits are on the bright side of things. Most of my callibration tables show at least some hits as much as 0.7 or even more above the dead average magnitude (call it every third or fourth star), and up to around that point, even when I don't hit the mark, I'm usually not that far from it. What it comes down to is my distances going 'up' are probably more accurate than his from 0.4 to around 0.7 above the dead average.

But Tsvetkov had only V to work with, which meant his accuracy rate probably fell off pretty dramatically around 0.4 going up. And he probably puzzled a bit over the super luminal stars, much as I have. Unlike me, though, he also did the Class III stars as well, and probably tried his numbers for those stars on the superluminals, which probably worked out much better. This, I think is almost exactly what he did with the second table of his in VizieR: rather than accept that there are quite a few Class V stars with absolute magnitudes a full point or more above dead average, he decided that those stars were class IV and III giants instead and the spectral classes were wrong. Given just how bright many of these superluminal stars are, and how contradictory spectral types can be...it might be possible to make a case for this. (I don't know enough, but it would explain a few things if he was right about this).

But that only accounted for the really bright stars. He still had quite a few stars that he couldn't legitimately call giants, or even subgiants, but were still brighter than his system could handle. At this point, he decided that it wasn't his system that was at fault, but rather the standard against which he was measuring that system; hence the accusation of systemic errors in the Hip. Now...I have come across the odd complaint about problems with the Hip before...at least the first version. If memory serves (and it might not) it was Gatewood (?) who after some careful parallax work of his own, raised questions about the error margin for some Hip stars, saying it was to 'narrow'. But...that was just a few stars, and he had his own parallax work to back him up, plus there has been the new reduction since then. So Tsvetkov has nothing going for him here.

Sound about right?

Sorry for running on like that, but it was percolating in my mind most of the day.

[ThinkerX]

Distance-dependent reddening correction? (once you know you're largely free of other systematic effects) I don't know what form that would specifically take...

I don't know either. The best notion I could come up with was - for stars past 100 parsecs - would be to set up a sort of 'sliding scale' where for every 50 (?) parsecs past that, I would assume that 0.1 (?) of the visual magnitude, at least, was lost to reddening, and adjust the distance moduli accordingly. I don't know if reddening affects just the V magnitude, or the J and H magnitudes as well. And yes, its an arbitrary fix at best. But I run out of stars a bit past 200 parsecs anyhow. So far, the alternative distances I've dug up - even those past 100 parsecs - are usually fairly close to mine (within 10-15% anyhow) about two thirds of the time...but I don't know if they corrected for reddening or not.

[ThinkerX]

Time for a bit of an update, I guess.

First off, the first 'block' - about 7700 stars - is nearly ready to go. How ready to go depends on how many more finishing touches I put on it.

Those 'finishing touches' would actually constitute a major project in and of themselves. The highlights:

The ASCC is missing something like 75% of the HD numbers it should have...and that is after I put several hundred back into the selection I've been mucking around with. It is also missing around 5-10% of the DM numbers it should have. Sorry, but I don't really have the time or the energy to put the rest of these numbers in.

There are two flags for double stars: the Tycho flag and the flag from the Proper Motion catalogue. Sometimes the two flags agree; very often, though, they don't.

The stars flagged from the Proper Motion catalogue usually led to 'real' double stars or multiple systems, many of them known for close to 200 years. For a few of them I've found actual orbits; for most of the rest - about 500 total - I usually found anywhere from two to six measures (Date, 360 degree Postion Angle, and Separation). About half these appear to be optical pairs or wide common proper motion stars; with the remainder being binaries. In maybe fifty or sixty cases, where I managed to dig up 4-6 measures taken at long enough intervals, it might be possible to plot rough orbits.

It is a different story for the stars with just the Tycho flag - usually there is no additional information available for these systems at all, even in the Tycho Double Star catalogue, which *really* should have them listed, at the very least. I actually suspect some sort of systemic error with the catalogue here.

I have yet to take a truly detailed look - and I might not bother - but it appears the Tycho flag for variables is about as reliable as the Tycho flag for double stars.

I did do a proper motion cut to try to weed out giants, subgiants, and superluminious stars. I deleted only a very few, and settled for flagging the rest - about 600 total - as suspect. This gets to be difficult past about 150 parsecs because the proper motions of all the stars are greatly reduced, and the giants are not moving radically slower than the dwarfs.

[t00fri]

The ASCC is missing something like 75% of the HD numbers it should have...and that is after I put several hundred back into the selection I've been mucking around with. It is also missing around 5-10% of the DM numbers it should have. Sorry, but I don't really have the time or the energy to put the rest of these numbers in.

This task is really straightforward and not time consuming at all, provided the right tools are used: i.e. Perl

With Perl, one simply reads both the ASCC and a suitable HD cross-index catalogue with J2000 coordinates into memory and then searches for coordinate matches within a small circle around the desired ASCC star coordinates. For each hit, Perl prints out the desired HD numbers along with the input data in any custom table format. This way, the job is completed quickly without chances for misprints... Not more than 1/2 page of Perl scripting!

Depending on the cross-index catalogue used, other star properties may be checked during the same task as well.

Fridger

[ThinkerX]

Double, double..

Got the initial processing of the second big batch of stars taken care of...over a week ago. Just over 18,000 stars total so far, from Dec +89 down to 0. Because of the way the ASCC numbers are working out and because there are more stars in the southern declinations, I am still figuring on 40,000 stars or a bit over for this catalogue.

But it is the double and multiple stars which has been occupying the bulk of my time and attention lately. Just finished with the basic researching on not quite 600 stars flagged as multiples (the Proper Motion Catalogue flag, for whatever reason, seems to be much more accurate than the Tycho flag). It looks like around 50 of those stars - at least according to Simbad - were false positives. Another 100 - 150 (bit hard to tell sometimes) look like Linear or Optical systems. The rest seem evenly split between wide Common Proper Motion pairs (talking separations on the order of hundreds to thousands of astronomical units, possibly a light year or two in some instances) and close binaries, with orbital periods on the order of a couple decades to a couple centuries. Only a couple dozen of the systems I looked up had actual published orbits; I think I dug up enough measures (anywhere from four to eight) to make at least rough guesses for the orbits of another hundred or so.

Generally, what I'm trying to do here is to get enough measures to first determine if a double system is actually 'real' (two or three measures is enough to determine whether a system is Linear/Optical or CPM or possibly binary. However, while two or three measures *might* be enough to determine if a system is binary, it is not enough to make even a good guess at the orbital period and parameters. For that...well...you *might* be able to do it with four measures, if they were spread out enough, but six or eight are better, so that is what I tried to aim for.

But when I think about it a bit...

18,000 systems. By most estimates, at least 3500 of those should be double or multiple star systems. Yet...even if I accept the Tycho flags as valid (which I'm leery of, because aside from the flags themselves, I can usually find no measures or even other evidence of duplicity for those systems) and count the Linear/Optical systems as doubles, I still have only about a third that number - around 1400 or so. Discount the unsupported Tycho Flagged systems and the opticals...and its probably fewer than 1200. Which means that either:

1) there are a whole bunch of double systems that some how missed getting counted; or

2) there are vastly fewer double star systems out there than the estimates usually allow for.

Also...there are simply a huge number of multiple systems - at least a quarter of the ones I looked at - where there are very few measures, or where the last measures date from something like fifty or a hundred years ago. Time and time again, I would come across what initially looked like an interesting binary system, only to see it had a grand total of four or five measures (two or three of which I couldn't find) and the last good one was from like 1928 or 1977 or some such.

[t00fri]

It is straightforward to check your stars (RA+-delta, DEC+-delta) against the standard Wahington catalog of visual double stars (WDS) and the SB9 catalog (spectroscopic double stars).

Fridger

[ThinkerX]

It is straightforward to check your stars (RA+-delta, DEC+-delta) against the standard Wahington catalog of visual double stars (WDS) and the SB9 catalog (spectroscopic double stars).

Yep, the Washington Double Star Catalogue is the first source I check. They give the first and last measures for a given system, and give a total as to how many measures there are...but so far as I can tell, there is no way, short of directly contacting them on a star by star basis, to get the rest of those measures.

The CCDM is the second source I check. About three times in five, I can get an intermediate measure from there, usually from 1893 or 1905 or 1914 or some such, which suites me just fine. I suspect most of these measures come from the olde 'Burnham Double Star' and 'Aitken Double Star' catalogues. Been trying to find online copies of the BDS and ADS without luck. (Rather strange, since ADS numbers are a common means of identifying double stars even now).

Past that...the actual papers that give double star measures (at least the online versions) start in the mid 1950's, and run up through the early 1990's. (These are not in VizieR). I have to go into ADS and look them up each and every time. I am contemplating digging a bit deeper into the ADS system here.

Actual online catalogues of double stars start around 1995.