I have been interested for some time in the paper by Joe Garlitz, “Using unfiltered images to perform standard filter band photometry”, JAAVSO 45, 2017, 75-85. He references earlier work by Arne Henden and others on a similar theme.
Garlitz used an SBIG CCD camera. The transforms in his Figure 2 show that the native properties of the sensor approximate those of an Rc filter. I repeated the experiment (in less detail) using a ZWO ASI294MM CMOS camera. The results (see the attached plot titled Unfiltered Transforms ZWO ASI294MM, displaying the same type of data as Garlitz’ Figure 2) show that the native properties of the sensor approximate those of a Johnson V filter. If the V transform plot is assumed to be linear (which I believe it is not), the slope (coefficient) is 0.017, nearly a horizontal line. The coefficients for the other filter bands are much larger.
The main problem with transformed photometry on unfiltered images is that the colour index of the variable star cannot be measured from those images and must be obtained from another source. If the colour index of the variable changes enough during a cycle, and particularly if transformation coefficients are ‘relatively large’, calculated transformed magnitudes will not be accurate.
Results from the ASI294MM camera for the V passband do not suffer much from these problems. V-iMag (iMag is the unfiltered instrumental magnitude of a star) is minimally affected by colour index. The coefficient from the (assumed linear) transformation is small, and could for some purposes be ignored, the observer simply calculating non-transformed magnitudes. The results of such an option are illustrated in the attached graph titled Target V Mag Err v. Target/Comp B-V Difference. The error for target/comp B-V differences up to about 1.0 magnitude is about +/- 0.02.
All of this is relevant to me because I have a short focal length (200mm) small aperture (71mm) system, imaging through an f/2.8 Canon camera lens. Using unfiltered images, time series photometry of eclipsing binaries for determining times of mid eclipse would be possible on fainter targets than those imaged through a V filter.
Roy
In the previous post, it was noted that the plot of V-iMag versus B-V was not linear. In fact the relationship is described by a second order polynomial expression as in the attached graph headed V-iMag v. B-V. The same can be said for B-iMag v. B-V and Rc-iMag v. B-V.
The slope of the V-iMag curve varies with B-V. When the latter is zero, the slope is – 0.278. At B-V 0.8 the curve is nearly flat, with a slope of -0.009. At B-V 1.2 the slope is 0.152.
The transformation coefficients for B-V values well away from 0.8 are obviously not small. I therefore suggest that, if unfiltered images from the ZWO ASI294MM are used for the calculation of non-transformed magnitudes, comparison stars should be chosen so that their colour indices are as close as possible to the colour index of the variable star.
Roy
Hi Roy,
While the AAVSO does allow users to submit un-filtered photometric data, it must be stressed that such observations often have less scientific value than their filtered counterparts. With that said, there are a few exceptions.
The most common use cases for un-filtered photometry are within the fields of exoplanets and cataclysmic variables. In the case of exoplanets, the primary way we contribute to the field is by observing potential eclipses to refine orbits and transit times. This works because exoplanet transits are effectively gray to a few-percent level and uncertainties in instrumental bandpasses don’t cause too much uncertainty in eclipse timing. In the case of CVs, much of the interesting astrophysics occurs at very short timescales so researchers prefer short high time cadence, high SNR observations. In this situation, unfiltered observations make a lot of sense.
Generally speaking, I would advise against applying transforms as described above because they fail in a multitude of situations, particularly for objects that exhibit large temperature variations.
If you choose to acquire un-filtered data, please reduce them in the standard way against either a V or R photometric reference frame. To submit the data, encode them as either CV (clear reduced against V) or CR (clear reduced against R).
Kind regards,
Brian
Hi Brian,
I agree completely with almost everything you wrote.
The reasons I trialled the transforms on unfiltered images were, first, simple curiosity about what they would look like with my equipment and second, if the results were favourable, photometry on unfiltered images would allow me to observe fainter eclipsing binaries than I do with a V filter.
I should explain that I do not aim to derive transformed magnitudes from unfiltered images: I wish to derive non-transformed magnitudes reduced using comp star V magnitudes. The results imply that this can be done successefully (see attached graph, which has more data that the similar graph in my first post). The range of B-V values for the stars was 0.44 to 1.554.
Your wrote: “I would advise against applying transforms as described above because they fail in a multitude of situations, particularly for objects that exhibit large temperature variations.”
I agree, and this was what I was referring to in my initial post when I wrote: “If the colour index of the variable changes enough during a cycle, and particularly if transformation coefficients are ‘relatively large’, calculated transformed magnitudes will not be accurate.”
In summary, any results from the method as described would represent CV magnitudes (clear/unfiltered reduced against V). The targets will be EW eclipsing binaries, which vary little in B-V across the cycle. The results of the trial as described above suggest that the photometry would be sufficiently accurate to yield valid time series photometry for eclipse timing. It is submitted that this is another exception to the rule that unfiltered observations often have less scientific value than their filtered counterparts.
Roy
Hi Roy,
Indeed. I think you’ve done a good job selecting a science subject (eclipse timing) and targets that are well suited to this technique.
Before you spend time observing, I would also suggest that you check the photometry from TESS to see if they have collected data on the stars in your list. TESS has a broad bandpass ranging from 600 - 1,100 nm that deliver extremely precise photometry with uncertainties well below 3% for most stars (their lower limit is 60 ppm). This would be extremely difficult to beat from the ground. The primary downside to TESS is that its pixels subtend some 21", so many stars can be blended. Obtaining broadband photometry on stars not observed by TESS or those that are blended could be quite useful.
Brian
Hi Brian,
The target stars are from the EB database of Variable Stars South, and there are about 220 of them. We undertook a project during the past year of checking all of them for TESS data, and added eclipse timings from TESS, when available, to the VSS database. I have already observed several dozen of the VSS EB stars with my own equipment, obtaining non-transformed V magnitudes.
Further testing of unfiltered images of the sequnce of 12 stars in the transform plot in the first post shows that there is a limitation, in that the most accurate unfiltered photometry is obtained only for targets and comps with B-V colour indices from about 0.5 to 1.2.
Roy
