Sun, 03/08/2015 - 21:59
Hello! I performed my first I,B,V measurements of ASAS J174600-2321.3 using ensemble photometry and MPO Canopus/Photored. The error range is about 0.25 magnitudes for all 3 filters.
The curious thing is that the error ranges for PNV J17032620-3504140 are 0,04 to 0.12 magnitudes, and the two fields are close together.
What would be the best way to sort through the cause of the higher error range of the ASAS variable? Could some of the comps be off? Best regards.
Mike
Hello! I've tried different apertures, check, and comps, but I cannot seem to get the error under 0.25 mags. I would appreciate suggestions on what to consider in order to lower the magnitude uncertainty. Best regards.
Mike
Mike,
What is the SNR of the target star in your images?
What exposure time are you using?
Are you stacking images?
Phil
Mike,
My suggestion is don't do ensemble. Rather, pick a specific C1 and C2. (I'm using star 119 (000-BLP-263) as C1 and 123 (000-BLP-264) as C2 . This allows professionals to adjust the data if they conclude adjustment is needed in the 119/123 magnitudes. But they have a firm base to work from. With ensemble they can't make such an adjustment.
If those of us measuring ASAS J14600-2321.2 use the same comp stars, the data will all be consistent which makes their job easier.
Gordon
Hello! I am stacking two 60 second images. The SNR in I is 98 andn FWHM is 4.14. When Canopus matches the chart, the star is at 17:46:00.13 and -23:21:15.8, though I doubt such miniscue difference from the published coordinates matters.
For ensemble work, Canopus/Photored usees the stars in about 15 arcminute area around the variable - about 15 stars. Best regards.
Mike
Mike,
There is an AAVSO CCD sequence for this star. For starters, I'd suggest you use those for your ensemble. Just use VSP to plot a printable CCD chart with your FOV, then replot and check the box for the sequence.
Using the AAVSO sequence should removed some of the possible problems, variables disguised as comps and poor accuracy in comp star magnitudes. You might also consider concentrating on just the V filter for a while until you get the problem figured out. At a SNR of about 100, I agree that you certainly should be getting better results. That said, I'd also suggest you pump up the SNR a bit, at least for now, by stacking more images.
I'm not familiar with the photometry tool in Canopus. What measuring aperture are you using? Are you sure none of your comps have been saturating? Are you confident that your flat frames are okay?
Phil
Mike:
Would an image be worth a 1000 words in this case? Can you take some print screens of the image in Canopus with target/comp/check shown, and a table of the applicable data? Share them with us?
Ken
Good idea Ken. In addition, the FITS of the image inself might also help.
Phil
Thanks. I'll upload the files.
I'm using the check/comps from the AAVSO CCD table for the FOV of my system, about 14 x 18 arcminutes and there are 17 comps for this FOV in the table.
Darks and flats seem fine. As I mentioned in my first post, the PNV variable is close to the ASAS variable, yet the errors in the magnitudes are much less for the PNV variable. Additionally, the other stars, both LPVs and short period variables (like SZ HYA) that I image during the runs have very good error range, and are consistent with my past imaging sessions. Perhaps I am mistaken, but I would have expected larger mag error ranges for the other variables during a run if there was a systemic problem with my equipment and calibration that has appeared.
One comp was saturated. When I eliminated it, the mag error range was still about 0.25mags in all three filters - BVI
The field is very full of stars, especially in the I filter. Might there be a faint star close to the target in the aperture that affects the error range?
7 of the 17 comps are fainter than 15 in V. Perhaps limit the comps to those with mags brighter than 15?
Finally, many of the I mags have errors of of over 0.2. Might this be part of the problem? Perhaps limiting the photometry comps as suggested above would be best instead of performing ensemble photometry? Best regards
Mike
You're right, the large I comp star errors contribute to the problem. I assume the error calculation uses the normal square root of (the sum of the individual star errors squared/n). So exclusive of the SNR induced errors, you are getting a major contribution caused by the comp star errors. (I looked at the errors of all the comp stars up to 150, and combining them (22 comp stars) contributes an error of 0.21 before you even look at the SNR portion of the error.) We could check with the professionals, but I know in the cataclysmic variable world Joe Patterson prefers a single comp star so he knows exactly how the measurments are calcuated. He can adjust them later if he concludes the reference magnitude needs adjustment. And it really helps if the different observers use the same comp stars.
So I suggest just one comp star. I chose 119 because if its proximity to the target, inate brightness improving SNR, and low B-V. It does have a large I error, but if most people use it the professionals can go back later and get a more precise reference magnitude if they need a more precise magnitude measurment.
We may need to switch to a dimmer comp star during the eclipse as the target dims.
Gordon
When you are having problems it is always good to simplify.
Why don't you reduce the images using a single comp star and a single check star. Choose a stars that are as close as possible to the target and are about the same magnitude as the target star. Also examine each to insure they are in the clear as much as possible or at least you can choose an aperture, gap and annulus that is appropriate for all three stars (comp, check and target). By appropriate, I mean that the aperture is the appropriate size and the annulus is clear of companion stars and/or background stars. Also choose a comp star that has a reasonable error.
Also I would suggest working with one filter at a time. A good place to start is the V filter.
Now you can do some detective work. You may find that this improves your error. In any event, it will make what's happening more obvious. You can calculate K-C. It should be a straight line with an error similar to what you would expect from your target star. If I'm having problems I like to take a rolling 5 point standard deviation of K-C (assuming a time series) to see if the error is changing through out the night.
You can also use apertures of different sizes and see if that changes your error.
If you are using 17 comps in a crowded field, chances are that several will have annulus contaminated by companion stars or background stars.
Good luck
Jim Jones, JJI
Hi Mike,
The major difference here is the magnitude limit. For PNV J17032620-3504140, the comps go down to V=14.5; for ASAS J174600-2321.3, they extend down to V=17. Both fields are crowded (near the Bulge; in fact, the ASAS variable is in an OGLE field, which is why so many variables show up on the plot).
Since the nova was 9th magnitude at maximum, the Sequence Team created a sequence as bright as they could go, but only down a few magnitudes, with the probable intent of revisiting the field later when the nova faded and extending the sequence fainter. For the ASAS variable, the minimum magnitude was known, so the sequence team tried to find stars as faint as the variable's minimum. By doing so, however, they had to use stars down to the faint limit in the APASS catalog, where the photometric errors are large (especially in such crowded fields). These faint stars are not necessary for CCD photometry since the sensors have a linear response, so removing all comparison stars that are fainter than, say, 15th magnitude will undoubtedly give you better results. Note that there are two sources of error here - the catalog errors (which are large near the catalog limit), and the photometric errors from your image. Since the ASAS variable has S/N=98 according to your post, and it is currently about V=12.3, the comps that are at 16-17mag are probably barely detectable in your image and with very low signal/noise (and shouldn't be used as comps, even in an ensemble). Both of these sources of error make the calculated error for your ASAS variable larger.
Second, note that the Ic errors are larger for the ASAS comps than for the PNV comps, even at the same magnitude. This implies that some erroneous image was included in the APASS catalog for Ic for the ASAS field (which is ~20 degrees away and which was calibrated on different nights). When building APASS, I'm currently just looking at V and removing entire datasets if the V magnitude for a field is discrepant. On occasion, V will be ok, but one of the other filters taken that night will be discrepant, and passes through to the catalog. The final catalog will be selective and delete single frames instead of sets, so things like this Ic error for the ASAS field will be caught. The final catalog will also do a better job on faint stars in crowded fields. However, you have to work with the catalog currently in existence, not some future improvement.
Bottom line: use the 8 comps brighter than V=15, and I bet your BV mags will be good. Ic will also be good, but may have a systematic offset from the standard system. However, everyone using this sequence will have the same systematic offset, so you will agree with any other observer also using the sequence.
I hope this helps!
Arne
For quite some time I had promoted the concept of helping new observers come into VSO, by first starting out as a visual observer, then "moving up" to CCD. But, it has become very apparent that the knowledge, principles, equipment required for good CCD photometry is so different, so much more detailed and "advanced" than visual photometry, that I think this idea is not really practical. Visual is not much of a "stepping stone" for CCD at all.
So, it appears that new observers should decide at the outset, whether they want to take the visual or CCD path, based on their own interests and abilities. The AAVSO should gear their introduction to beginning observers, to highlight the plusses and minuses of visual and CCD separately, and suggest they adopt one path or the other.
Mike LMK
Hello! I had a significant reduction in error when using fewer stars for last night's run.
When I used the complete CCD table with 23 stars for enseble photometry, the error ranges were I = 0.387, B = 0.147, V = 0.184.
When using the first 7 brightest stars in the CCD table, the erros were I = 0.144, B = 0.055, V =0.027.
Perhaps the table should be trimmed to limit the comps with high error ranges? Best regards.
Mike
Hi Mike,
Good to see that your errors reduced by using only the bright stars. Before I'd trim the VSP comparison-star list, I'd first look at your signal/noise for those fainter stars. My guess is that their table photometry is ok, but your image photometry has low signal/noise for them, and that is why your errors increase when including the full table.
Arne