Return to Home Page
From John Greaves: === Right then, just to cover it, and as it has probably all got a bit confusing (well, I'm a bit confused so why shouldn't everybody else be): UCAC2 and CMC13 are ASTROMETRIC SURVEYS They are not PHOTOMETRIC SURVEYS, nor are they in anyway photometric in their remit. They weren't intended to do photometry either. I never, ever, said they were photometric. I mentioned this in my first mail. I said the authors of them mentioned this in their documentation, and gave urls as reference. So, why did I bang on about the magnitudes in them? Simple. Despite the fact that they are ASTROMETRIC SURVEYS, they have columns of numbers in them, and some of these columns have the description "magnitude" appended to them. And I know full well what happens then. Nigh on a decade ago around a million BT and VT "photometric" values were published. They were instantly used as Johnson B and V. Still are sometimes. When folk started converting them to Johnson like they should do, they used the simplest equation. The documentation itself contains over a dozen different conversions dependent on BT-VT and flags etc, some even not being mathematical, being simply lookup tables. But people used them anyway as if they were high quality, usually all the way down to the catalogue faint limit, and ignoring photometric quality and variability and duplicity flags. That's what happens. In a year or two Arne Henden's TASS camera based V-Ic should be out, and eventually Pojmanski's ASAS3 stuff should be usable, and hopefully published as a source catalogue too. Then it'll all be academic (well, except for those few regularly observing 17th mag stars...) Till then, magnitudes are used where they can be grabbed. It happens. Folk do it, whether right, wrong or indifferent. UCAC, hitherto mostly ignored in these climes as it only went up to about -23 degrees declination is now at the tender mercies of northerners. CMC13 has sneaked out, and although an astrometric survey, they bothered to put a real filter in the front of the CCD. Well, an SDSS r' one. I did a graph or two to highlight the potential problems. To keep this as terse as possible, I'll use Berto Monard's mail as a template to give short comments on relevant points, as it does bring up common problems and to some extent, self convictions not necessarily based on anything. Some of this some of you will already know, but this is trying to give ideas on how to think on the data, and there will be points not widely realised. Brian and Arne have already made two strong points re UCAC photometry in vsnet-chat, for instance. You can always disagree with me, coz I will give places and ways for you to look yourself. Berto Monard opined : > I read with interest about the tests of J Greaves, his conclusion and personal > recommendations, followed by comments by others. > > I wonder in how far UCAC2 is different from UCAC1...... Brian noted differences here: http://vsnet.kusastro.kyoto-u.ac.jp/vsnet/Mail/chat6000/msg00739.html > To derive V magnitudes from UCAC is indeed more tricky since O-V is much larger > (about 3 times) than O-R (O magnitudes from UCAC), Aha. I had thought CCDO was so called "broadband V" Wratten orange filter in front of a CCD stuff. Apparently it is also CCD stuff done against UCAC red. Wonder if CCDOs in archives are mixes of these, then. Quite different too. > so you end up with a 3x > larger conversion error. The accuracy of the conversions is proportional to the > accuracy of the knowledge of the colour term of the stars in question. Well, I'm going to have to look up and see what the spectral response graphs for these passbands look like. This makes it sound like one long square wave. > What do > you use for that: USNO-A2.0 or perhaps USO-B1.0, what else? How accurate is > that? and how precise between different fields? USNO B1.0 is probably good enough to tell whether a star is blue/white, yellow or red. Then still, if 0.556(USNO B1.0 B1-R1) = 0.6, it could be 0.3 or 0.9 in the worst case, ie white, yellow or red! See http://vsnet.aerith.net/astro/color_conversion/JG/USNO-B1.0.html IMPORTANT. USNO A2.0 red and blue colours contain no provenance! Use of USNO B1.0 red and blue colours usually teaches what the situation with these colours really is. In the North, B1 and R1 are nigh on contemporaneous O and E plate images, respectively. B2 and R2 are J and F plates of random epoch. J and O are blue, but J has a filter in front of it, O doesn't. (O is IIIaO is a plate emulsion, and has nothing to do with yellow filters and/or UCAC red magnitudes, by the by). In the South not only are B1 and R1 noncontemporaneous, they are not necessarily on the same 'scopes, mountains or even on the same continents, as B2 and R2 are. B1 and R1 seem to be AAO and B2 and R2 ESO. Except for when they aren't. If there are two lots of AAOR for instance, and no ESOR, R2 may be mentioned using the second AAOR, or only R2 may be mentioned. It looks to be arbitrary to me. But then again, USNO B1.0 is an ASTROMETRIC catalogue. So, okay, you can sort this out, you say. Thing is though, next to nodody uses USNO B1.0 info, they use USNO Ax.0 info off of their CDR copies!!! And these only contain one each of a red and blue magnitude. And these are taken from the self same plates as USNO B1.0 uses. So, you can see the point, you don't know what mix and match of red an blue you have in the South when you use Ax.0. Okay, the response comes, if the star is constant it don't matter? You'd think. However, once you start using USNO B1.0 colour from time to time, you notice something. Although B1 and B2 may be quite close to each other, sometimes R1 and R2 look to be about half a mag to a mag different, for an otherwise constant star! That'll be coz R1 is AAO red and R2 is ESO red. Why red seems more prone than blue to this, I dunno. (NB, I know this is somewhat anecdotal. Anyone feel free to plot it out). This is no doubt why when Sebastian had a bash at emulating my Northern B1-R1 messing about with Southern B-R, he got errors up to +/- 0.5 in derived B-V. And that's gonna be the same for USNO Ax.0 red minus blue. Contextual point : you try to select comparison stars as near as possible in colour as to your variable, or at least in general within the range of B-V 0.0 to 1.0. With an error of +/- 0.5 you can estimate if your comparator is within an acceptable range, but the colour correction you can apply to any mag becomes meaningless. > Despite what is in the readme.txt, UCAC is based on 'filtered' measurements of > the real sky, not of archival photographic plates and therefore much more true > than the data in USNO-A, B-1,2.. How can someone say or hint that USNO-A, B > could be possibly better. UCAC is not filtered. Don't know what 'filtered' means, but if it means what I think it means, the eyeball is 'filtered'. Then again, I've far more respect for visual measures than a lot of professionals have. I've seen far better visual stuff than stuff coming out of a professional CCDed scope and true filter at times. But CCD "photometry" is another issue, and one of ooruri- chan's hobby horses (he's far better qualified to comment on it than I, and has done so). > Referring to the large numbers of test stars in John's tests, there could by no > means be a distinction based on colour terms. It's all in one pot. Plots of red-red were done against colour (essentially colour-colour diagrams), as mentioned. Any colour term would be meaningless, as the scatter was large to obscenely immense. That goes for UCAC too. > That's not > what you do in a comparison sequence. Here you select stars of the proper > colour before you work on them. At least that's how IMO sequences need to be > made. That's obvious. Some of use have been gawking at 2MASS J-H (Ks can be prone to nonstellar excess). There's a helluva spread twixt B-V against J-H, and B-V against J-Ks, and even J-H against J-Ks for that matter! Let alone any single colour against these. (V-Ic versus J-Ks looks tightest so far, though doglegged). Sure you could get a nice colour term out of it, and if you never showed anyone the plots, they'd never know how large the scatter was. I don't just shove the r^2 correlation coefficient onto the graphs just coz the spreadshit has it as an option, yer know. I'm lucky it is an available option, else I'd have to struggle to work it out! I wanna know how good these automated fits software and programs make are. Not just quote the damn things as given and as spat out, regardless. Why am I wingeing on about J-H? Well, if you're going to apply a colour term, you're going to have to get a colour from somewhere. If there was a barrelload of B-V available, we wouldn't be considering all this messing with obscure magnitudes in the first place. And no, USNO red and blue don't give colours of any real worth. Above. > I have quite some experience with UCAC1 and made several R comparison sequences > which have stood good comparison to later dedicated field photometry. > Therefore my Q above since John tested using UCAC2 material. Nonquantified statement. Experience shows me that odds and ends often look okay, you've seen a plot against only a handful or Arne photometry I sent out last night. Bad ones are often forgotten. And what does "stood comparison" mean? To 0.5, 0.3, 0.1, 0.05, of a mag, or what. Problems may go away if 'ensemble photometry' is used. But most folk are going around using one to two comparators at most. No smoothing available there to speak of. > I have no experience with ASAS nor TASS magnitudes but I can hardly see > soundness in using the latter above magnitude 12.5. I still believe that Tycho > and Hipparchos give good B and V magnitudes for up to 11.5V. It's to about V = 8 for Hp and V = 10.5 for Tycho2, not 11.5, let alone 12.5. I'll go dig the url to the gifs out... http://ftp.lowell.edu/pub/bas/BVT2mBVL.gif http://ftp.lowell.edu/pub/bas/BVTL_VJL.gif http://ftp.lowell.edu/pub/bas/VJT2mVJL.gif http://ftp.lowell.edu/pub/bas/tycho2.readme done by some bloke called Greaves, back in 2000 apparently (according to the date stamps). I don't want to be told Bessell will fix this. I'd accept being shown how much Bessell _improves_ on this. With _big_ samples from large or multiple areas. I believe some professional and analytical look-sees have been done wrt to this matter since that earliest of quickie assessments shown in the above urls. I love the ADS, try adsabs.harvard.edu. > Perhaps ASAS can > fill a gap up to magnitude 13.. 12.5 would be safer, though as Sebastian has noted, you can get a reasonable mean down to 13 or so. > Properly converted UCAC1 magnitudes can IMO be used up to magnitude 15 (from 11 > upwards), assumed a proper colour selection is done. No doubt. Given a nonlinear solution, acceptance of and allowance for trends, and a proper colour correction from somewhere, and inherent scatter taken on board as internal error afore full error is assessed. tain't done though. > The expected uncertainty > for 15 magnitude stars, derived that way is better than 0.2 magnitude in R, > about 0.5 in V. Have no illusions, the other database will not give you > anything better in accuracy, although you might think so. Where the hummer do these numbers come from? Folk may feel I witter on some, and graph things to death, but that's so I can show folk the workings, where it all came from, what the assumptions are, whatever. Invariably I get telled stuff, like this, that I've to accept someones on faith. r'_CMT can be damn tight, surprisingly so. There is no colour term to speak of against g'-r' (or was it z'-g'? can't remember) as shown in its documentation. There is a CTE problem in RA, serious attempts were made to compensate for it, again documented. r'_CMT has been assessed by the CMC13 authors. > Precision and > accuracy are not the same. Too damn right. The amount of crap carrying +/0 0.05 errors or better is legion. > For deriving R magnitudes above 12R, as for use in > unfiltered photometry, UCAC1 is far better a dataset than the others. Dunno. CMC13 ain't been tested in usage yet. TASS Rc is better between 12 and 13 incidentally, just of limited sky coverage. > Back to the test range for John, possibly using UCAC1 values this time and > calibrated fields, and red ears for Mike and Sebastian..... ;-)] Now, if he'd done me a plot of all his UCAC1 derived R mags against later photometric Rc provided from a good source, like Arne, I wouldn't be so wound up by that last sentence. There's some good stuff in TASS Mk IV, and some odd bits of rubbish too. The release is a prerelease release, much the same way as for ASAS3. I've had reports of good agreement for TASS MkIV V in comparison to sequences based on comparison with same object Landolt, Stanton and Henden photometry. The TASS folks are working on tightening up raw data selection of Tom Droege's stuff, and Michael Richmond has declared an intent to from-the-bottom-up make the whole thing more rigorous sufficient to make a reference catalogue good to the +/- 0.03 level in V, and though not necessarily as tight in Ic (???), good enough for quality colour information. Not sure of the details, just stuff I've picked up off of the lists. Now, to compare such efforts in direct photometric surveys to a bunch of unfiltered bonus magnitudes from an astrometric catalogue in an unfavourable light is a bit damn weird, dontya know? TASS and ASAS3 can be reprocessed and improved, especially the former due to it having two colour information. Still not sure whether ASAS3 has implicit Ic, or what. And in general, a word of warning to all. There ain't no "thing" called TASS, even though there is. When Arne's catalogue comes out, it'll be a TASS Mk IV camera survey, and may or may not contain the word TASS in its title. Folk will thenceforth start having to have their head wrapped around what exactly they mean when they say "TASS", ie whether it be Arne's stuff, Tom's stuff, one of the innumerable Michaels' stuff (I call it xMASS sometimes, where x is the unknown number and MASS is Michaels' All Sky Surveys, which is a bit unfair on Tom). I'll answer specific worries folk might have if they like, if I can, but I ought to drop this thread now as folk will be getting bored with it. It's an important matter in some ways, but as is usually the case on such matters, the less concrete information there actual is on a subject, the more that it is possible to talk about it. Roll on the time it becomes irrelevant. Cheers John
Return to the Powerful Daisaku Nogami
vsnet-adm@kusastro.kyoto-u.ac.jp
