Affiliation
Vereniging Voor Sterrenkunde, Werkgroep Veranderlijke Sterren (Belgium) (VVS)
Wed, 01/30/2013 - 09:30
After checking this star in VSX, I found out that its type was changed to "non-variable". It is constant in ASAS-3, Hipparcos and GCPD data. I hope I don't have too many of these stars om my observinglist! :-)
It is nice to see that, except for two outliers (6.1 and 6.5) all my observations of this star are between magnitude 6.2 and 6.4.
Maybe two possible improvements for WebObs:
- a warning if a dated chart is used, this gives me the opportunity to update it.
- a warning if a star is a non-variable
Wow! I've been observing this star for years. Wasn't aware it's no longer considered a variable, wonder if I should keep observing it?
My observations were between 6.4 and 6.6. Thanks (I think) for the news.
Yes, it doesn't makse sense to keep observing it.
The recent photometric surveys show it has been constant for years (1989 to 993 and 2000 to 2009 at least!)
ASAS-3 light curve (disregard the points before JD 2453000, they are saturated):
http://www.astrouw.edu.pl/cgi-asas/asas_variable/053020+0412.3,asas3,0,0,900,300,0
HIPPARCOS light curve:
http://cdsarc.u-strasbg.fr/viz-bin/vizExec/Vgraph?cat=I%2F239%2F.%2F25785&-graph=0%09&P=0&-graph=-Y%09mag&-x0=-&-x1=-&--bitmap=900x300&-y0=6.1&-y1=6.6&-y2=-
If there is any variation it is below 0.02 mag.
Cheers,
Sebastian
It's interesting, but sad, to note that we have over 10,000 observations of this star. If the average visual observation takes a couple minutes and you add another couple minutes to that for either entering the observation into WebObs or submitting by mail in the old days, we've wasted about 667 hours of people time on this star just to find out it is boring.
Drop this from your program immediately. Don't waste any more precious time on CK Ori.
PASP, 85, 42 (1973) "Some variables of spectral type K" by Olin Eggen
"CK Orionis: Although several visual observers have noted erratic variations up to 0.5 mag, this object probably does not vary appreciably."
1973 - 40 years ago ...
Cheers,
Doug
[quote=SXN]
It's interesting, but sad, to note that we have over 10,000 observations of this star. If the average visual observation takes a couple minutes and you add another couple minutes to that for either entering the observation into WebObs or submitting by mail in the old days, we've wasted about 667 hours of people time on this star just to find out it is boring.
Drop this from your program immediately. Don't waste any more precious time on CK Ori.
[/quote]
Once you look past the "knee-jerk" reactions regarding the "wasted effort" on CK Ori, it becomes apparent we have an extraordinary valuable set of data here. Think about it, 10,000+ observations by hundreds of observers over half a century of what everyone thought was a variable, but is in fact constant. Basically, a perfect "blind" experiment on observer technique!
A real bonanza for those who wish to mine data on the long or short term accuracy of visual or CCD observers. With such a large number of points, one can study practically anything - Visual vs. CCD accuracy, individual observer accuracy, effects of comp stars on accuracy, long term changes in observer accuracy, etc., etc.
No need to setup complex experiments and tests, we have the answers right in front of us now.
Just eyeballing the 50 years of data, it shows some remarkable things:
1. Range of visual is over a full magnitude, range of CCDV is about half of that.
2. The accuracy of a few selected visual observers is much less than the overall spread, typically 0.2 magnitude, but some are less some are more.
3. Overall slope of the data is about zero, which for such a large number of points implies it is in fact a long-term constant star.
I would recommend this data set be saved as a special case that anyone can mine and analyze for answering some of our fundamental questions on observer accuracy.
Mike LMK
I agree with Mike that this is an interesting test case.
I think there is a psychological factor lying behind the observed variations of both this star and T CMa.
One expects the star to vary...
I also observed some small amplitude variability in CK Ori 144 obervations between 1998 and 2003 with a mean V= 6.23 (0.02 mag. fainter than the true value) but ranging between 6.15 and 6.36 (usually 6.15-6.29). So this shows a ~0.1 mag. error in those binocular observations.
The case of T CMa also showed variability of the 9th mag. star this time between V= 9.11 and V= 9.31 with a mean V= 9.24 (0.06 mag. fainter than the true value) over 30 observations between 2000 and 2002. The comp star magnitudes used for those observations were from Tycho and were probably slightly off so a zero point comparison with the other case is not worthwhile.
But again there is a 0.1 mag. spread. But the worst thing is that there seems to be some up and down trend on my "light curve".
I think that psychological factor is playing a key role here. I have a lot of other stars that are CST with much better results but they weren't as famous and publicly recognized as variables like T CMa and CK Ori!! :(.
Also I always expect much better results for binocular stars than for telescopic stars. I feel more comfortable estimating binocular variables due to the larger field of view and the better quality of the image I get with my 7x50s than with my eyepieces.
My mean binocular accuracy is 0.07 mag. So CK Ori was worse than that. The orange color and the mentioned psychological factor were surely playing a role (also you improve your techniques over the years).
The results for blue stars and those having blue comp stars and also good regarding magnitude (similr to the variable) and position (close to the variable) can be even better than 0.05 mag.
So, yes these are good cases to test one's skills and to learn from your own mistakes in order to aim for the higher acuracy you can get.
Chris, about your comment on the long term variability of CK Ori, sorry that I don't follow you on this one but you are assuming that the observed variability is real and it is hard for me to agree with that.
We need to admit some of the limitations in visual observations. And this is one of them.
I get very angry when I see my T CMa light curve but I know that all these observations are useless :((. And I try to remember how they were made, and why I may have got those results in order not to make the same mistakes again.
Cheers,
Sebastian
One more thing, this JAAVSO paper by Peter Williams already pointed out the confusion on T CMa's identification back in 2006:
http://adsabs.harvard.edu/abs/2006JAVSO..34..235W
It will be used as the reference for the VSX updates.
Cheers,
Sebastian
And here is yet another "useful" data set: IU Ori http://www.aavso.org/lcg/plot?auid=000-BFS-626&starname=IU+Ori&lastdays=21000&start=&stop=&obscode=&obscode_symbol=2&grid=on&visual=on&v=on&unknown=on&unvalidated=on&fainterthan=on&pointsize=1&width=600&height=425&mag1=&mag2=&mean=&vmean=, more than 4000 visual observations of another constant star. This one was also found to be constant already in 1973 in the same paper by Eggen that Doug mentioned before.
Patrick
I'm new to AAVSO but numbers and stats were "my thing" so yes, I agree, the large sample size of observations makes this data very useful for anyone wanting to use this as a case study for visual observation accuracy studies. Can we assume the margin of error would be similar to all observations of all stars? Granted, the magnitude and colors of the stars would make a difference from some "standard error range" but still it shows some base to observer inaccuracy.
Question: again I am new ... we all can understand a visual magnitude error range but why would the Johnson B have a range of 6.33 (rounded) to 6.17 -- and this is discarding outliers on both ends -- from a light curve I ran from 1 JAN 2015 to today, n=101?
Since the software used is mostly the same, or only a few major software programs used, is this variation in photometry normal? Is it explained by camera quality differences more than the software? Failure to do calibration? Once I gain experience, I want to do DSLR photometry. If CK Ori is a constant, would one expect this much variance in magnitude reporting from photometry?
Thank you in advance; as a new member interested in photometry, the variance reported from a "constant" star is interesting.
Bob S
Hi all
This sad and scandalous expose is useful. 'Why?' is a question I ask myself for every star, and every night. For instance, years ago I did 'CST' for a season because no one else was looking, and I wanted to be sure. There was a question. Bully for me. But I dropped it after the season when it became clear I was wasting my time.
My program changes all the time; new targets fitted in, with hopefully a smaller number being dropped. We need to keep an eye (and brains) on the ball. Why do I observe this target?
Best to all. alan.
A very teachable (and learnable) moment. It demonstrates how difficult it is to stay on top of all the info, and that necessary info isn't always transmitted as efficiently as we would like. It's also a valuable lesson in the necessity to question the status quo!
KL
Hi Bruno,
This may help to identify constant stars in your observing program. If you log in to VSX and have entered your observer code in your profile (if not go to http://www.aavso.org/vsx/index.php?view=account.top to do it), you have the possibility to look for the stars you have observed. Choose this option from the special searches box at the top of the VSX Search form http://www.aavso.org/vsx/index.php?view=search.top.
Patrick
It gets you to wonder how many other variables we observe that aren't variables.
Rich (TYS)
T CMa was another example of a more or less famous variable that was actually constant.
The worst thing is that it was wrongly identified as a different star in the old AAVSO charts.
This is T CMa's light curve using the LCG.
You can see how the star that was observed at the early epochs is a bright 9th mag. star.
In fact, it is TYC 6541-1249-1 with V= 9.18 and no variability according to ASAS-3:
http://www.astrouw.edu.pl/cgi-asas/asas_variable/072121-2528.0,asas3,309,-26308.0000,500,0,0
Between 2000 and 2008 observations for this star and an 11th mag. fainter star coexisted, probably due to the use of different charts. The 11th mag. star is actually T CMa according to the GCVS position.
From 2008 onwards, only the right star was being observed, surely due to the fact that the charts were made using VSP and VSP takes the star's position from VSX which in turn takes it from the GCVS.
BUT, the 11th star is not an SR variable with a period of 309 d. and a range 9.0 - 11.0 V. It doesn't vary at all and it has V= 11.4. It has nearby 12 and 14th mag. companions inside a 18" radius. and the combined light curve of the trio (V= 10.8) according to ASAS-3 looks like this:
http://www.astrouw.edu.pl/cgi-asas/asas_variable/072126-2526.7,asas3,309,-26308.0000,500,0,0
The star's B-V is 0.1 so it is not a red semirregular.
There is no bright large amplitude variable in the area.
You made me remember this case. We haven't acted upon it over these years.
I will be revising the VSX entry that still shows it as an SR variable and we will correct the observations in the AID moving the bright data to the right star.
I wasted time on both the wrong bright T CMa and CK Ori in the 90s and early 00s.
Cheers,
Sebastian
Thanks for pointing this out, very nice feature! And it also shows the total number of observations.
I luckily had no more surprices :-).
Back in the 1970's I can remember CK having a good .4 mag variation. I still observe the star regularly, but agree the variation is much less. Wouldn't it be a good idea to keep observing the star to see if the range continues to slowly decrease or if it starts to slowly increase again? This could be some sort of long term cycle.
Chris Stephan SET
Robert Clyde Observatory
Lake Placid, Florida
Just so that everybody knows: UW Dra is constant according to Hipparcos data. There are 4196 observation in the database, from which 6 are mine. Luckily my variation range was limited, my estimations were all magnitude 7.6 or 7.7. :-)
It would be nice if the AAVSO could publish a list of variables that are now considered constant. Rich (TYS)
Rich, we're working on that now actually. I hope to post something within a day or so. I've already worked up a spreadsheet showing variables with no appreciable visual amplitude that I will also post.
Matthew
That would be cool. Rich (TYS)
Hi,
The following page is now available. It was created using VSX to find stars with a Variable Type of "CST" or "CST:". There may be additional stars that are genuinely constant but not catalogued as such.
Visually Observed Stars known to be Constant or not Measurably Variable
As I have time I'm going to go through a list of stars where the standard deviation of all visual observations is smaller than 0.5 or so. I'm finding that some but not all of these stars might also be "constant" to within measurement errors. That would also be a good community project: run Fourier transforms of all such visual light curves and flag those where the only detectable periods are around one month or one year. There are a lot of those.
Also, if you haven't seen it, John Percy et al have recently done studies of a number of visually observed stars and found objects that could safely be dropped. One such paper was this one in JAAVSO from 2011.
Matthew
I would second Bruno's (BBI) suggestion that a bold warning on the specific VSP chart be given: "This star is know to be constant, please do not observe." An independent list of constant stars may not be enough to discourage observations. Kevin Paxson - PKV
I've emailed that suggestion to Will McMain to ask how long it would take to implement in VSP.
Hi, Kevin,
Since VSP takes the information on the variable star from VSX, the VSP header will already say that it is CST if it is catalogued as such.
That is, if you make a chart using the star's name. If you want VSP to add a warning message every time a CST star lies in your chart's field of view, I can't see how that can be implemented.
Cheers,
Sebastian
Is it possible to add validations in WebObs? One that gives a warning, not an error, when an estimate for a CST star is being added to the database?
Hi Matt, all
I was interested in Matt's comment:
I'm sure it's occurred to you that we could make use of VStar (DCDFT in particular) for that.
The Mean time between observtions plugin could also be useful.
How best to coordinate this activity, which I would be interested in participating in?
David
How's that for a project acronym?
More seriously, yes, it would be a useful project and VStar would be a good choice for it, too. What I'd like to see is just this:
If the answer to question 2 is "no" in both sets of tests, then there probably isn't anything periodic there. That's not an absolute criterion for whether or not to observe them, since stars don't have to be periodic to do interesting things. However, if it's not periodic and there isn't anything else interesting going on, then visual observers' time is more likely to produce valuable information if spent on other stars. You can make the "nothing else interesting" criterion more rigorous, like a standard deviation of 0.5 magnitudes or less of the entire data set. But just running and looking at the results of a Fourier transform would be useful.
I can make a list of stars with rms values of 0.5 visual magnitudes or less. Anybody want to take the lead on the analysis? It'd make a nice JAAVSO paper at least....
Matt
Nice acronym Matt.
I am interested to see whether there's any signal, other than monthly or annual as you say, in such observations.
What else could be useful here? Self-correlation analysis (e.g. in the Percy and Terziev paper you cited), ANOVA?
I would like to play an active role in this effort. I can attempt the Fourier analysis you request.
Equally, I'd be happy to make this a group effort; key to that would be to set out a clear procedure, elaborating upon your steps starting with a subset of the stars in the list.
If you think someone better qualified in data analysis should take the lead, feel free to say so. I would be happy to take this on at least initially however.
David
Actually David, I think you'd be perfect to head the analysis, and I think as developer of VStar you're probably more qualified than most.
I've attached a list of stars that I think might be good to investigate, where the cut was made entirely based on a MySQL query "stdev(magnitude) <= 0.3". I should caution (a) 0.3 is very arbitrary, and (b) there are some things I know are visibly variable despite having low amplitude. As an example of (b), do a Fourier transform of the visual data of SY Mus -- the signal around 620 days is far above the background noise despite having a semiamplitude of only 0.2 visual magnitudes. SY Mus is an excellent example of why visual observations *ARE* worthwhile! Going down the list I see a bunch of other objects that I know for a fact are variable, so this might actually be a good test of both the data and the analysis.
The two criteria I described above are arbitrary, so I encourage you and anyone else interested to do something more rigorous. And regarding the methods, I think using multiple methods does have a place. Fourier and ANOVA are nice when there is a strictly periodic signal present, but something like autocorrelation or self-correlation can pick up things that aren't so strict (like an RV Tauri).
Just to be absolutely clear to people casually reading this thread, the list I've attached is a list of stars that would be interesting to analyze. This is NOT a list of stars to be dropped!
Matt
Thanks Matt.
I'll make a start on a subset of those objects, make notes regarding approach, and report back here or in whatever context you think makes sense. Others can get involved to reduce the load of course.
I can feel a self-correlation plugin coming on...
I've read about it in John Percy's book and seen it applied to example data but need to take a closer look at it.
One thing you get from VStar AID dataset load is the known period if any exists. Something that might be useful is to repeat that SQL query but include the recorded period as well. That way I can do an initial automated AID data load and phase plot generation with VStar scripting. The previews generated will be useful in seeing whether periodicity exists, permitting the less obvious cases to be targetted first with DCDFT etc.
Another approach is to deliberately not look at the recorded period and just rely upon inspection, DCDFT from scratch for each.
What's your view?
Your example of SY Mus was a good one. Visual inspection indicates some periodicity. Magnifying and using the mean time between selections plugin gives a period of ~620 days.
DCDFT (standard scan) confirms this:
A phase plot at the top hit (f=0.0016, p=1/f=622.826) gives:
Regards,
David
Hi Matt
Ignore my comment about another query to include the period. VStar gets that from VSX (of course) not AID.
Sorry for any confusion my comments may have caused.
David
Hi David,
No problem. I think I prefer the suggestion you hinted at, namely not giving the expected period. I think it's a useful exercise to look at the data first, and prior results second.
Regarding prioritization of that long list, I think the stars that are constant (CST), untyped (VAR or blank), and anything uncertain (:) would be useful to do first. There are so many variables in there that I know are variable, they're more useful as test cases -- I'd be surprised if the period didn't show up like it did with SY Mus. Even if there's no clear period, the variability is clear for some of these things despite the rms of the data being small. Visual observers really do a great job, even on some weakly variable stars.
Matt
Hi Matt
Agreed. I'll proceed on that basis. Thanks!
David
Hello Matthew and David.
I am very interested in Project “LAVA”.
I have done some analysis using VSTAR on four stars listed in the spreadsheet that Matthew attached to an earlier post, to prove to myself how well DC-DFT can reveal periods. A quick summary of the “test case” results are :-
eta Gem; the period found from visual data was the same as that from V-Band data, despite the ‘scatter’ in the raw visual data been over twice the amplitude of the stars light variation. Also with the visual data the ‘true’ signal stands out at a higher level than the ‘one year’ alias signals given by falias = freal ± n/365.25, whereas with the V-Band data, (as there is less of it), the ‘true signal’ and alias signals where at the about same power level.
R Lyr; a DC-DFT ‘Standard’ search of visual data, shows low frequency signals at a much higher level than the correct signal, though a signal corresponding to a period of ~ 46 d is visible above the noise. A DC-DFT period range search of 10 d to 500 d shows much more clearly what is going on, the signal at 46.4 d is at about the same level as a 260 d signal, (the 260 d signal is probably due to the observations been concentrated over 9 mths when R Lyr can be observed at a reasonable time i.e before midnight). Also as well as the ‘one year’ signal it is easy to see a signal at ~ 29.5 d, showing that the ‘Lunar Cylce’ has effect the visual observations. Also present is a signal peak at ~ 65 d that could be a 'one year' alias. VSX lists the period for R Lyr as 46: and mentions a period of 64 d. The data from two individual observers shows the period as 46.4 d.
eta Aql, straight away a DC-DFT ‘standard scan’ showed a signal at around the correct frequency from the visual data. A further high resolution ( 0.0001 d) search showed the period to within 0.002 d of the period listed in VSX.
X Sgr, a DC-DFT ‘standard scan’ produced a misleading result and suggested a period of ~ 20d. Narrowing it down using Period range searches and doing a search in the range of 1 to 10 d res 0.0001 showed a period that agreed with the one listed in VSX to within 0.002 d.
I’m personally impressed that ‘inaccurate’ visual data can reveal a Star’s period to an accuracy of a few minutes!
The visual data from individual observers may or may not reveal a correct period. The results from one observer produced a good result for both eta Gem, and R Lyr. Whereas another observer’s data didn’t produce a good result for either star.
As well at the “Test Cases” I have so far had a quick look at most of the stars listed in the ‘spreadsheet’ as “CST” or “CST:” the only star to show anything interesting other than the ‘one year’ signals is :-
LU Del, There is evidence for a period of ~ 95 d, from all the visual data. After finding that period I looked at the reference given by VSX to IBVS 3186, the star was “discovered” variable with an amplitude of 0.2 Vmag and a period of 100 d. Also the spectral type of LU Del is listed as G8III, so as it is a 'coolish' Giant, I think I’m correct in saying that does suggest it is reaching the point in its life, where it will pulsate. So perhaps LU Del is a case where a high number of visual obs can show the period of a very low amplitude star.
As for CK Ori, the strongest mid to high frequency signal is one for around 30 d due to the ‘Lunar Cycle’ affecting the visual estimates!
Great stuff Pete. Sorry for the delay in replying. I'm at a conference for the 2nd half of this week so am fully occupied mentally by that at the moment.
I will properly respond on the weekend at the latest.
I will probably move/re-start this thread in a different forum given the subject matter we're getting into now.
Thanks.
David
Sebastian has looked at the ASAS data and found no variations above 0.02 mag amplitude in LU Del. There may still be a periodicity in the recorded magnitudes in the database, but this could be due to a sequence star being variable and not the target itself.
G8III stars that are variable are more likely to be RS CVn type than pulsators.
Cheers,
Doug
Thanks Doug,
It may be a coincidence that a star in the sequence used has a period that is close to the one found when LU Del was discovered variable. There is defiantly evidence for a period of 95 d in the visual data for LU Del. I know it seems unlikely that low precision visual data can pick up very low amplitude variations, but when a lot of data is all averaged together.
I apologize for my limited knowledge of variable stars types. If LU Del is a RS CVn type star, could 100 d be the eclipse period or would a period of that order be far too high ?
Attached is a word document, showing the VSTAR analysis results for LU Del.
Pete
Hi, Pete,
>>> HR 7923 (HD 197249) was announced as variable in IBVS 3186, where the period is given as 100 d and amplitude of 0.2 Vmag. The low amplitude signal with a period of 95.92 d, found from the visual data covering a date range of 20.8 yr, closely matches the 100 d period.
However, from IBVS 3186:
JD-2440000 delta V
6762.200 -0.601
6763.192 -0.478
6766.308 -0.435
6768.308 -0.618
... which is completely inconsistent with a 100 d. period. The full amplitude of variability is seen between consecutive observations 1 and 2 days apart. So these observations are just scatter to me.
An one isolated V obervation at V= 6.36 with an error of 0.08 mag. is also meaningless. The zero point may also be wrong.
There are no variations between 1989 and 1993 and between 2003 and 2009 so I wouldn't trust the visual results on this one.
>>> If LU Del is a RS CVn type star, could 100 d be the eclipse period or would a period of that order be far too high ?
A 100 d. period is possible (on the high side) but being RS not necessarily means the system is eclipsing. Most are not, showing only a rotational modulation caused by spots. But there is not even a mean magnitude change here between the HIPPARCOS and ASAS-3 data. Only the visual observations show variability and I don't think it is possible to detect such a small amplitude variability, specially by combining data from different observers with different color responses.
Best wishes,
Sebastian
I notice in some of the earlier posts that the late Olin Eggen is quoted as stating that one or other supposed variable is in fact constant. It might be borne in mind that it was also Olin who (in a reference I do not have to hand) stated that S Per had only a small amplitude, and poured scorn on the English amateur who had asserted otherwise. I have great respect for Olin's contribution to astronomical thinking, but he is not the final authority on the occurrence of variability in a star. He himself generally made only very few observations of any particular star.
Tom.