Transformation Coefficients and M67.
A transformation coefficient generally tells how your instrumental value differs from the true value as a function of star color. This is usually a small correction; important if you need high accuracy but unimportant if your target and comparison stars are similar in color. Hendon, Arne, 26 July 2007.
Frankly, I generally do not apply transformation coefficients to my observations unless there is a specific request for a specific target, self compelling or if I am running alternating filters (usually V & I) doing a time series.
I am also not implying, herein, that M67 is the only or the best option for imaging as a Standards Field for your own computations. There are other suitable FOV’s.
I just finished up my annual computations from three nights of imaging M67 and thought I would share some information that might be of benefit for others. When I say annual, it is true that I try to do this every year, but in this year’s case they are the first for my recent replacement scope.
In past years I have used the chart and data (Arne’s) located at:
http://binaries.boulder.swri.edu/binaries/fields/m67.html
Awhile back, Arne uploaded some revised and expanded M67 data to the VSD so that we can obtain a Standard Field Chart of that data from the VSP:
Enter the coordinates: RA 08:51:24 + 11:45:00 and then select Yes, at the bottom of the VSP chart options, Would You Like A Standard Field Chart?
The field is quite crowded with photometry taking these standards as faint as ~ magnitude 16 and contains many duplicated values (which is OK for this purpose).
However, It can be a challenge, therefore, trying to use this Standard Field Chart, because of the crowding, and my first suggestion would be to reduce the number of standards by limiting the magnitude to say 13; but that still leaves a number of duplicates which almost requires that you plate solve your own images or, as an alternative, use Aladin (down load to your desktop) to help you identify specific comps:
I choose to instead use the older chart for identifying the VSP’s standard stars as it is quite visible with its original number system and makes it easy to locate the comps when you are analyzing the data; I of course used the updated BVRI data from Standards Sequence. Once you examine the first links data you will catch on to what I am describing.
Choosing which standard stars to use and how many is, from my perspective, a bit of an art form as much as it is science.
I have made it a habit through the years to zero point on the number 1 star in the original published data and then to subtract all the following instrumental magnitudes from the standard values. If I find differences greater than about .08 I simply do not use that standard star. The twenty stars that I ended up selecting this year had standard deviations of the Standard value vs the zero pointed instrumental value of between .015 & .023, depending upon the selected filter.
The other issue is mixing of color choices (B-V) that go into your selections. In my case the average B-V value was .634 with a median of .582. This average seems to fit well with my typical targets.
Here is the B-V data for each standard star that I choose.
-0.062
1.264
1.109
0.575
1.093
1.135
0.096
0.219
0.128
1.076
0.606
0.290
1.073
0.405
0.456
1.000
0.670
0.561
0.589
0.568
So, as an FYI, I will share with you which specific 20 standard stars I used for my computations, with the first number being the matching star # from following link:
http://binaries.boulder.swri.edu/binaries/fields/m67.html
while the second data column is obviously the standard star AUID from the M67 Standards Field.
1 000-BLG-879
2 000-BLG-886
3 000-BLG-887
4 000-BLG-888
5 000-BLG-889
7 000-BLG-891
10 000-BLG-892
11 000-BLG-893
13 000-BLG-895
14 000-BLG-896
15 000-BLG-897
16 000-BLG-898
17 000-BLG-899
20 000-BLG-901
23 000-BLG-903
24 000-BLG-904
25 000-BLG-905
31 000-BLG-910
33 000-BLG-911
37 000-BLG-914
I am not suggesting that you use these specific standards or this many or few; Just a starting point for your own research if you choose to image M67 for TC computations.
For those of you who are curious or would like to compare your own TC’s (and please keep in mind that they will vary a lot from system to system) here are my BVRI computed values:
Tbv = 1.008
Tvr = 1.091
Tri = 0.965
Tv = -0.005
Tr = -0.049
If you compute TC’s for two or more filters the formula’s can be found in the Sarty paper:
http://www.aavso.org/sites/default/files/Transforms-Sarty.pdf
Here is my data for the two cases where I may just use two filters:
V & I filter: Tvi = 1.020
B & I filter: Tb = 0.003
Computing the Transformation Coefficients
I just read this morning in anther forum post, that Arne wrote they have a volunteer writing a program for computing your transformation coefficients as other progress towards making the process more user friendly. That is great news for many observers:
http://www.aavso.org/stage-two
In the meantime, however, I really like the Pricilla Benson paper as it lays out the whole process of how to generate your own transformation coefficients in a simple and straightforward manner and has enabled a lot of us to generate our own spread sheets for both the creation of the transformation coefficients as well as their application to observations.
http://www.aavso.org/sites/default/files/benson.pdf
There are other reference pdf’s that the AAVSO has available on this topic and here is another one:
http://www.aavso.org/sites/default/files/ccdcoeff.pdf
Arne’s suggestions, to me, some years back, were, when following the Benson paper or any of the others, with similar suggestions, would be to:
1) Instead of plotting (R-r) vs (R-I) for Tr to instead plot (R-r) Vs (V-I)
2) Instead of plotting ((V-v) vs (V-R) for Tv to instead plot (V-v) vs (V-I)
V-cI has more change with star color and so results in a better transformation; V-Ic is also more sensitive to interstellar extinction; using Rc-Ic means you use Rc, a bandpass that has H alpha in it; which is a very prominent line that can be in emission or absorption causing an error in transformation, especially with novae, where H alpha emission can dominate the continuum in the Rc bandpass. Arne
I have continued to follow this advice.
I hope some of this information may prove to be useful for your own efforts.
Ad Astra,
Tim Crawford, CTX
Thanks Tim,
Lot's of great information there, the Benson paper especially, as you say lays out the whole process in a straight forward manner, I already feel I have a grasp on the subject.
Many thanks again,
Douglas.
Tim,
Thank you for that great post! I will take you up on that offer t go over my M67 data in more detail. The next month or so is very busy , but you will hear from me soon after that....
Ooops; Actually it looks like you didn't really offer that. You just provided some of your data for comparison. I didn't remember it correctly when I read it quickly yesterday.
Thank you for the great post and information. This will help me a great deal!
Thanks for
Tim,
Thanks for sharing
Mike
So now that I have added a corrector to my Dahl Kirkham telescope I am recalculating transformations. This time using M67. Is there any particular reason that I shouldn't simply pick the first 30 stars in Arne's original sequence, excluding variables and suspected variables? These appear to be the 30 brightest ones. I use a smaller than normal aperture size when measuring cluster stars per Arne's recommendation in another string on transformations. Therefore, crowding doesn't seem to be an issue for more than a couple of these stars (like #22 and #12. #11, #18 and #30 seem to have sufficient separation). For M67 an aperture radius of about 5.25 arcsec (10.5 arsec diameter) seems to work well. In my images this amounts to radius of about 1.7 FWHM (diameter 3.4 x FWHM) I also do a two-pass regression. I calculate a regression for all the measured stars. Then I eliminate stars with residuals from the regression line outside the 98% confidence interval and run the regression a second time. From past experience on NGC7790 most frequently no measurements are rejected and even for relatively noisy data rejection of more than 1 measurement out of 30 is rare.
So, Anyone know why I shouldn't just pick the non varying stars in the first 30?
I don't know what aperture size Arne used to calibrate the cluster in his original PDF sequence and its relationship to the image FWHM. Can anyone supply that information?
I don't pick a single star as a comp for the others. My software allows me to use an ensemble approach. I measure the raw magnitudes of all the selected stars calculate a zero point for the image based on the average of all of the offsets and then add the zero point to the individual raw magnitudes to get instrumental magnitudes. So the "effective color" of my comparison is the average color of all the stars I measure. This probably doesn't do significantly better than just picking one of the standards with B-V in the range of 0.5 to 0.7.
---------------------------------------------------------------------------------------------
IGNORE THE FOLLOWING PARAGRAPH: As Arne points out below, the "color bias" from choosing a specific star affects the Y intercept of the regression lines. It doesn't affect the slope and the coefficients are the slope or are calculated from the slope. The Y intercept doesn't matter for this purpose. Star #1 is bright and uncrowded and would be a good choice if positioned correctly for the part of the field you image; so my appologies to Tim. WBY
If I were using a single comp to measure the others, I would not choose #1 because it is quite blue compared to most of the standards in the cluster. I think #4 (B-V = 0.576) would be better choices. The numbering is from Arne's old PDF M67 finder chart.
-------------------------------------------------------------------------------------------
Brad Walter, WBY
Hi Brad,
Picking the first 30 is fine. What I tend to do is make sure that a couple of very red stars and very blue stars are included. The majority of stars in any field will have "average" color, so you have to work a little harder to get the extreme ranges.
Remember that the zeropoint is just some arbitrary constant - it doesn't matter whether it is a number you created out of thin air, or the value of a single star, or of an ensemble, as you are not using that zeropoint to calibrate any other field. It just creates an offset for the least squares line, and you are only interested in the slope of that line. So don't go overboard worrying about that aspect.
Arne
Arne, your absolutely right about the zero point. I use the residuals between the instrumental mag and the standard mag as a quick check to see if I have made a mistake in my photometry take-offs (as in selecting the wrong star or the aperture being out of position) or have a problem in a particular image (an unfortunately positioned cosmic ray hit, for example). Also, I rarely have just one image. I essentially always have at least two in each filter because I take images ithrough multiple filters in pairs, IRVBBVRI for example, so that the average airmass of all colors is essentially the same. If the offsets from standard magnitude for a star are very different in the two images, that also gives indication that I have a problem with that star in at least one of the images. If I have the time I try to do this initial check while the synthetic apertures are still in place on the images. If I have made an error,I can correct it easily and get the revised data quickly. If you only have a couple of stars on a couple of images it doesn't save much time, but if you have to reset 30 apertures on several images that are shifted, it saves quite a bit of time and eliminates a lot of swearing. I don't align the images because the software I use for photometry does very precise centroid alignment - even if just shifting images without ratation or scaling. That means the alignment is sub pixel and flux gets shifted between adjacent pixels. I could do a single star alignment in Maxim first without using centroids. That simply shifts images by full pixel values without any reallocation of flux among pixels. I'll have to do that next time I have a bunch of shifted images. Now let's see; which bit of string around which finger is supposed to remind me to do that?
Brad Walter, WBY
Hello All
I just wanted to alert our observers that M67 is up in the evening for taking your Transformation Images. For me, M67 kind of sneaks up on me, as I think of it as a spring object, and then, with the time change, its gone, and I have missed it. I observed my regular program for several hours, and then M67 was really nice and high before the merridian. Even if you are not planning on doing the reduction, having the images available will make it more compelling.
Speaking of Tools to calculate and apply your transformations, I used PTGP 5.5, now renamed TG (Transform Generator) to determine them. I used VPHOT to measure the stacked images in BVRI. Then loaded the new coeficients into TA 2.29, along with the error bars. It took a little refresher with the Vphot videos, and the TG help file to format everything properly. Having the M67 Standard Star File Catelog as an option in VPHOT was great. It really helps. Once familiar with these Tools, it should take no more than 10 minutes to get your coef and to apply them. I will have to admit that it took me several hours to get thru this, even though this is the second time for me. I am not a regular user of Vphot, so it took some time to reduce the data, and I was watching TV at the time (Multiplexing).
I wanted to give a shout out to George (TA), Gier (Vphot), Gordon (TG) and Arne (M67 Stds) for making this process so seamless. Give it a try.
WGR
Gary
Hi all,
To calculate the transform coefficients insist some efforts but it works well and I have done for my VRI filter set. To transform separate photometric measurements is easy task with VPhot but what I need are some hints how to transform time series photometry data sets. I normally use time series observations with hundreds of images and I wonder whether is necessary to transform all that data.
And something more, I use preferably Sloan type filters and I need to know whether exist some g’r’i’z’ standard field acceptable to use for transforms.
Velimir
Hello Velimar and All
I have two suggestions on transformations. FIrst for time series, I use TA, Transform Applier, an excellent tool developed by George Silvis and team. Its available from the home page: Data>Data Analysis>Transform. ALso available from there is TG (PTGP) a program to determine your coeficients.
I also suggest using VPHOT to make your measurements. Select the Standards Stars from Catelog as your sequence, and VPHOT will find all these stars and measure then, without having to point and click each one, and it deals with the crowded fields. It outputs a file that can be inputted to TG (Transform Generator) and it all works very nicely. I just did it after a 3 month hiatis (sp?). TG also plots each coef and you can point and click and delete discrepant points and it plots the new BFSL in another color so you and compare. You can add back the point if you wish.
Once you have your stacked images, and you become familiar with TG, it should take no more than 5-10 minutes to get your coeficients. It normally takes me less than 1 minute to transform a night's time series, once I have the coef. I use the coef for 6 months to one year. I did them in November with NGC7790 and wanted to compare with M67. BTW: NGC 7790 also has its standard values in the catelog with M67. The coef agreed well.
Clear Skies--M67 is up nicely this time of year--go get your images.
WGR, Gary
[quote=PVEA]
Hi all,
To calculate the transform coefficients insist some efforts but it works well and I have done for my VRI filter set. To transform separate photometric measurements is easy task with VPhot but what I need are some hints how to transform time series photometry data sets. I normally use time series observations with hundreds of images and I wonder whether is necessary to transform all that data.
And something more, I use preferably Sloan type filters and I need to know whether exist some g’r’i’z’ standard field acceptable to use for transforms.
Velimir
[/quote]
TransformApplier can handle timeseries data. Gary has run hundreds of observations through TA, no problem.
If you would like to send me a copy of one of your WebObs submissions I can advise you if there are any issues. These might be how TA is forming the transform groups and fetching the reference data. Also include your current transform coefficients.
Cheers
George
SGEO@GASilvis.net
Hi Gary and George,
Thank you for advises and for the hints. First of all I will try to do some exercises with TA and TG to get familiar with the software. My transform coefficients for VRIc are only roughly estimated in the field of NGC 7790. I will do this once again but carefully in the M67 field.
Can you recommend me also the star fields suitable for Sloan g’r’i’z’ transformation?
Velimir
http://binaries.boulder.swri.edu/binaries/fields/m67.html
does not work
See earlier forum messages. This is not the source you want to use anyway, if you are trying to generate transform coeffs? If not, excuse my response.
Use magnitudes given in the "Standard Fields" option provided in VSP. Use RA/Dec of M67 or ev cnc to open this chart or photometry table.
Ken
Hi Ken,
Thank you for the information but actually I have no trouble to find and use M67 Standard field and a Photometry Table. In the APASS I found lots of stars with Sloan g’r’i’ data but I just wander whether they will appear as standard sequence in M67 frames or I have to create my own sequence manualy. I have no M67 frames to check it out now but if the weather is mercifully I will try this tonight.
Velimir
Velimir,
Yes, APASS provides BVgri mags. However, the M67 Standard field does not include the sloan magnitudes. Arne revised the Standard mags using J-C filters.
You could manually add APASS sloan magnitudes to the comps BUT they would NOT currently have the level of accuracy provided by the UBVRI mags. Arne has warned against using such data for transformations until a later release of APASS.
Ken
You may have found that link in the TG documentation. I attached a copy of the M67 field showing the old Boulder id's which you can still use in TG if you are manually selecting the standard stars and using AIP4WIN or MaxIm. TG will automatically go to VSP to get the most current values.
However, I strongly urge you use VPHOT to identify the standard stars. It will save you a LOT of time.