Skip to main content


7 posts / 0 new
Last post
rko's picture

I've been producing a couple of spreadsheets to transform CCD magnitudes in B & V using the equations given by Bruce Gary. It suddenly occured to me that in getting the transformed V mag, by subtracting the transformed B - v colour index from the transformed B mag, I'm using two images taken at different (though probably only slightly different) times. So what time do I use for the transformed V mag? I can see that with a slow variable, it probably doesn't really matter but what about a fast varying nova or Delta Scuti star?

Clear (un light poluted) skies

Keith Robiinson - RKO

HTY's picture
Midpoint of the V Image

Hopefully an expert will pop in here soon but I use the mid-exposure time of the V image to report the transformed V mag.  My reasoning is this:

The purpose of the B measurement as it relates to the V measurement is to produce a color index used to transform the V measurement to the standard system.  It will never be a "perfect" color index if the two images are taken at different times and the star is varying but hopefully it is close because the two images were taken very close in time to each other.  However, even though the B image was taken at a slightly different time, it does not alter the time that the V measurement was made so for the V image I would report the transformed V magnitude with the midpoint of the V image as the JD.

The same would apply to the B measurement.  I would report the transformed B magnitude with the mid-exposure point of the B image as the JD.

For a rapidly varying star, if you are taking a BVBVBVBV series, you could form the color index from the mean of the 2 B's immediately preceeding and following the V image if you think the color is changing quickly but I would still report that mid-exposure time of the V image as my JD for the V measurement.

Now wait for a real expert to appear and give the real answer. :P

....Tim (HTY)

rko's picture
Mid point of the V image

Thanks Tim. It seems to have the characteristics of a systematic error. I think you're right in using the mid point time and it goes without saying that in doing multi-colour photometry on a rapidly varying star, one has one's work cur out.

All the best


HTY's picture
Just to Clarify

Re-reading what I wrote, I realize I may have been more than a little verbose.  I did not mean to report the time as the midpoint between the B and V images.   

The V time should be reported as the time of the V image and the B time should be reported as the time of the B image.  

....Tim (HTY)

WBY's picture
Timing of images

I have Just a small variation on Tim's BVBVBVB sequence. If you use the same B image to transform different V images it is possible to establish a correlation between adjacent measurements because you are using the same B image in the transformation adjacent V images.  The variation is BVB-BVB-BVB or, for more "efficiency" in V, BVVB-BVVB-BVVB, which gives you two V images for every pair of B images. Extended to the full Johnson cousins filter set  the sequence becomes IRVBBVRI-IRVBBVRI... . I don't know how significant the correlation between images may be, but you may prefer to avoid the potential for it if you are making a light curve.

The data from the images pairs of the same color in each group are averaged. Different colors may have different exposures, but if you keep equal exposures for for a color pair and don't stop in the middle of a sequence group to re-focus, re-center, etc., then the effective time of the pairs all work out to be at the average of the midpoint times of the two B images of the full J-C series, or at the average of the midpoint times of the V images for the BVVB groupings.

When setting this up in Maxim as a sequence you have to differentiate between the first and second image of each color pair in the sequence setup. In Maxim 4, for example, you would setup two exposures for each filter in your sequence setup table with different suffixes, say I1, R1, V1, B1, B2, V2, R2, I2. Each image in each sequence must have a unique file name and setting different suffixes for the two exposures in each color accomplishes that.

Please excuse me if much of the detail is obvious or "old hat" information for you.

Brad Walter, WBY

Richard Wagner
Richard Wagner's picture
Timing/Sequence of Images

I'm working hard at getting my 'photometry pipeline' working and only just realized/remembered this issue of correlation between magnitude measurements.  I have always observed BVBVBVBV... (ad infinitum it seems like) and had been working towards using the average of the contrasting colour images on either side of each image to get the other half of the B-V.  So BVB gives me a single V and an averaged pair of Bs.  Step one image to the right and VBV gives a B and two Vs. Step one image to the right...  This results in correlations - each B image gets used to measure the colour for both the V images preceding and following it (and vice versa).  Is this important?  I could easily switch to using my images in groups of three with no sharing between groups: BVB,VBV,BVB,VBV.  On a few of my targets the exposure durations are significant compared to the period - say 5% for a full BVB triplet.  Is having more frequent more correlated data better or is less frequent uncorrelated data?  To what extent would this decision depend on the individual light curve (ones with bumps might prefer more data, while smoother curves could use the reduced correlation?) (And should I stop trying to observe stars that faint :-) ?)

Thanks (and sorry for dredging up an old thread)

Richard Wagner

HQA's picture
exposure timing

Hi Richard,

If you expect significant color change as a function of time, I'd suggest taking your data in the sequence VBBV VBBV VBBV

Then, for each set of 4 images, the mean time is identical for all filters.  Since B always seems to be the long exposure to reach a given signal/noise, I'd adjust the V-band exposures so that you get the same signal/noise as you do for B.

You are correct that doing BVBVBV and then getting colors by BVB,then stepping one image to the right and getting VBV results in correlated photometry (the same reason why boxcar averaging is uncorrelated while a running average is correlated).  I prefer to be uncorrelated, though in practice the amount of correlation is usually pretty small (color changes are far less dramatic than magnitude changes).


Log in to post comments
AAVSO 49 Bay State Rd. Cambridge, MA 02138 617-354-0484