single-filter transformations

Affiliation
American Association of Variable Star Observers (AAVSO)
Sun, 03/15/2015 - 17:58

From the AAVSO CCD Photometry Guide, Chapter 6:

Vvar = del V + Tv_bv * del (B-V) + Vcomp

where del means the difference between the target star (variable) and comp for the given quantity, and

del (B-V) = Tbv * del (b-v)

Where (b-v) is the instrumental, untransformed color.  So, in order to transform the variable's Vmag, you need to know the variable and comparison star color (the B-V quantity).  This is easily obtained if you have images in both B and V filters, as you use the instrumental color to derive the transformed color, and then use that to derive the transformed magnitude.  How large can this transformation adjustment be?  Assume that the V-magnitude coefficient is 0.2, and you are looking at a red variable (say, B-V = 2) with typical comparison stars (say, B-V = 0.8).  Then del (B-V) = 1.2, and the adjustment is 0.2 * 1.2 = 0.24 magnitudes.  Substantial!  That is why untransformed CCD magnitudes have so much scatter on red stars - each observer's photometry is very precise, but when combined with other observers, the accuracy can be poor.  This is why we ask observers to please consider doing transformations.

What if you only have a V filter?  How do you transform that data?  If the target star had a constant, known color, you could obtain that value from some catalog and use it in place of the derived color in the above equations to calculate a transformed V magnitude.  For many variables, assuming a constant color is reasonable:  a W UMa eclipsing star usually has nearly constant color throughout the cycle; a cataclysmic variable is usually blue; many low-amplitude variables like delta Scuti stars don't change much through their cycle.  In most cases, the cyclic color change can be considered a second-order effect:  you can get most of the way to the right answer by assuming a constant color.  In our example above, the variable might change from (B-V) = 1.8 to 2.2 through its cycle, which changes the correction from 0.20 to 0.28mag, so by assuming a constant 2.0 color, you change a 0.24mag offset into a +/- 0.04mag offset.  Not perfect, but you can see that the offset between observers has lessened.

This approach has been used extensively by the PEP group, where most observers are observing with just a Johnson V filter and we are assuming a constant, known color for the variable and its comparison star.  This again works because most of the stars on the PEP program are low amplitude stars, or at least stars that don't significantly change color through their cycle.

So I do recommend that you consider single-filter transformation of your data.  There are two issues to resolve.  First, you need that known color for the variable.  In an upcoming update to VSP, we will be using APASS mean colors for all of the variables that have been detected by that survey, and those would then be available for your use.  You can also look at the AID light curves, and where B and V are shown, calculate a mean (B-V) yourself.  Second, you need a version of TA that will work with known colors, rather than deriving them from two input filters.  That does not exist at this time, but we might consider an extra "flag" for TA to handle this single-filter option.

Arne

Affiliation
American Association of Variable Star Observers (AAVSO)
Single filter transformation and TA

When would you use single filter transformation? Since the process requries Tv_bv then the user must have both B and V filters from which they developed that coefficient. I guess you would use this when B is so dim that you can't get a decent exposure or you want a higher cadence for V.

TA could be modified to do single filter transformation:

You showed: Vvar = del V + Tv_bv * del (B-V) + Vcomp

instead of del(B-V) =  Tbv * del (b-v)  TA would use = (estimated(Bs-Vs) - (Bc-Vc)) since the estimation derived from APASS would be transformed colors. Getting that data from AID (the light curve generator) is a bit harder, but its an estimate: set the error accordingly.

TA could have another tab that would let the user catalog the estimated variable star colors and associated error of their favorite variable star targets. It could put the estimated color into the records' comment. 

Would you want a single filter transformation for the other filters: U, B, R and I?

B would work with Tb_bv and the B-V color.

U would work with Tu_ub and the U-B color

R would work with Tr_vi and the V-I color

I would work with Ti_vi and the V-I color

 George

 

Affiliation
American Association of Variable Star Observers (AAVSO)
VPhot Single Filter Transformation

Just a note that VPhot currently conducts single filter transformation when one enters the assumed/known color of the target in its sequence.

Additionally, this transformation can be conducted with an ensemble of comps. In this case each comp is used to calculate the transformed magnitude of the target from each comp and the final transformed target magnitude is then calculated from the average.

Attached are two print screens showing the assumed color entry page and the final magnitude ensemble photometry page.

Ken

Affiliation
American Association of Variable Star Observers (AAVSO)
How can I get the color index changed from N/A?

Ken - the VPHOT tool you cited is very useful - but I can't get it to work.

I  have a couple dozen obs of ASAS J174600-2321.3 that I haven't reported yet because it was requested that only transformed obs be submitted. VPHOT won't let me transform them even though I know the color (v-i) of the star. When I try to enter the color, there is an "N/A" listed in the color box on the form. It seems frozen. What do I have to do to change it?

Lew

Affiliation
American Association of Variable Star Observers (AAVSO)
Time Series Single Filter Transforms using two B & I data

As Arne pointed out in the beginning, single filter transforms should be possible on our data if the variable and comp stars have constant B-V and/or other color values.  What I've been doing on a few of my AM Herculis and CR Bootis time series is to shoot one or two B and I frames in the beginning of the sequence (say at air mass<1.6) and then shoot 250-400 15-second V filter shots, processing them in Transform Applier 2.38. 

Now how does do I get transformed 300 transformed V-filter shots if I only uses two B and I shots?  Simple, I cheat and DUPLICATE 298 B's and 298 I's run them through TA and once transformed, erase the duplicate B's andI's leaving only two I, two B and 300 V shots.  This process is very simple if one copies the MaximDL extended format into a spreadsheet and match the number of B & I data with the number of V data.

If I assume the variable and comp are not changing significantly during the run, is this proceedure "honest" enough with transform accuracy of the V-filter time series shots? 

I've started to shot additional I and B shots in the middle of the series.  A more refined part of this process would be to average-out B-V and I-V changes when taking BVI data in the beginning, middle, and end of the run.  This way if there is a change of color of the variable, an average can be taken in a process Arne alludes to in single filter transforming.

James

 

Affiliation
American Association of Variable Star Observers (AAVSO)
Single Filter transforms and TA

Greetings James,

Note that version 2.38 of TA will do the single filter transforms for you and save you all that record duplication. You have a number of options: determine the color at the beginning of the series, periodically through the night or insert averaged colors into the series.

This does almost the same computation as you describe above, but its not cheating! 

I've suggested in the past that transform groups be allowed to fill out by reusing observations. eg  BVRIRVB would reuse the I to make 2 4-filter groups or your case above where you reuse a B observation to fill in BV transform groups. Matt and Arne strongly deprecated this procedure. It creates a problem in AID because there the transformed records are expected to have all members of the transform group present, which in your case they will not.

The single-filter transform in TA documents in the comment portion of the resulting record that there was a single-filter transform and the color (and its error) that was used.

George

 

 

 

Affiliation
American Association of Variable Star Observers (AAVSO)
Single Filter transforms and TA

To:  George,

"This does almost the same computation as you describe above, but its not cheating!"

Thanks for the reply....now I feel a little less unsavory!   Upon thinking about this further, there will be a color change in a short period CV (Period<0.3 days), but unless we do a BVIBVIBVI.........sequence, it would be impossible to capture the B-V or I-V changes while doing long sequences of single filter time series observations.  An average B-V or I-V can be calculated using random or "shotgun" observations, but these might not correspond to the B-V color extremes when the variable is in quiescence or outburst.

I'll check out that feature of TA and continue to find ways/verify that any color change occuring to the variable is taken into account.

 

James

Affiliation
American Association of Variable Star Observers (AAVSO)
Xform with one filter

Hello James

Why not observe in 2 filters next time, and then you have it all.  It takes less time that "cheating".  If you already have the data, and are stuck, then you improvise as you have done.

I always observe in 2 filters at least.  

 

Gary

Affiliation
American Association of Variable Star Observers (AAVSO)
Xform with one filter

To:  Gary,

I use three filters actually (BIV), but I only use 2 of them (BI) for the 1st part of the run and conduct straight V-series images for the rest of the run; maybe 250-400 images, 15-seconds at 2x2 bin.   I know that alternating BVIBVIBVI......sequences woud be preferable, but AM Her changes so fast that it requires short cadence to catch all of its minute-by-minute variable behavior.  When I did asteroid photometry, BVRIBVRIBVRI....runs were done to catch the smooth rotational behavior of the object, but these short-period CVs seem more chaotic in their variability.

James