Looking at the light curve, I'm pretty proud of the visual observers. Your observations are mostly fitting into a peak-peak distribution of about 0.5mag, meaning the RMS uncertainty is in the 0.1-0.2mag range. Not bad!
I am NOT so pleased about the digital observers. Your observations are also mostly fitting into a peak-peak distribution of 0.5mag. Your distribution should be at most 0.05mag, or a factor of ten better. It is not equipment, as Ulisse Munari's ANS consortium in Italy can achieve 0.012mag total uncertainty between multiple observers, using amateur telescopes and CCDs. I think the difference is in technique. AAVSO observers are not transforming their data, are not observing with proper cadence and multiple image sets, are not watching for saturation or other defects in their images.
The professionals are NOT observing this nova - yet. It is still too bright for even a one-meter telescope; Swift is totally saturated with its imager, etc. It is up to us to create a quality light curve, one that shows all of the low-level bumps and wiggles. It is not just this nova, but any improvement we make here will also make subsequent novae observations better, and give better results for Jeno's project.
So what I am willing to do is work with each individual observer, but in the public forum. Let's examine images, consider how to obtain and use transformation coefficients, and learn techniques to improve the photometry and the reported uncertainties. It is an investment of my time and that of other photometrists that should reap rewards - if you are willing to suffer a small amount of public scrutiny of your data. Those who are AAVSO members can share their images through VPHOT; others can upload requested images to our ftp site where we can link to them for examples.
There are a couple of dozen digital observers of Nova Del 2013. Are you willing to help out?
I can now see a few observers with similar results. I posted JD 6581 in B. I also posted JD 6584 in B from a long run last night. The B was fainter again last night. I have simultaneous V time series, and the B and V agree nicely. The I data is not following suit. Its nice to see the PT and the Spec agree.
For those interested, here is a relatively deep V-band finding chart for the nova, taken with AAVSOnet TMO61 telescope with about 1.3arcsec seeing on 2013-10-12. The chart is 4x4arcmin with the nova centered. There are no nearby neighbors that will be a problem for a while. I'll take a deeper image in a few days to see what the nova will look like in quiescence.
V339 Del has undergone several updates to the sequence with the last being made on Oct. 17.
Updates occured to all values between Magnitude 8 & Magnitude 11 on charts 1 degree and less FOV.
Tim R Crawford, CTX
Thanks for pointing this out. I checked my data, and I was using the 80 for this night, and fortunately it did not change as a result of the update.
It does look like the 98 became .034 mag brighter as a result of the 17 Oct update. This would brighten any B measurements in the data base using the 98, and tend to agree with Robin/Buil spectra. We are so used to seeing B to be fainter than V in most variable objects because they are red, that its a little red flag to me. This is not your usual variable object, and may be more typical of Nova.
BTW: Great job by the sequence team. These new comps are going to be invaluable going forward on V339 Del. On JD 6584, I used a 4 star Ensemble of 121, 105, 125, & 122. Nicely spaced for apertures and sky annulea, and a good range of mags. We are also very lucky, in that the B-V's are all between 0.88 and 1.09. A little redder than the nova, but all very consistant so color effects should also be consistant.
I made a goof on the 2456571 data reduction: specified the compstar magnitude about 0.010 too dim. The data are fixed now, and now my measurements are slightly brighter than before (further distant from CCD values).
I put up some 2456584 data with rotten error bars, but I am still consistently brighter than CCD. I used two different delta B-V values based upon Gary's measurements that night. See the "notes".
Three recent (Oct 29.09) reported V values seem to show greater variation than seen previously:
My suspicion is that the larger number of available nearby comps in the target magnitude range leads to more selection options and points to greater target error due to potential errors in comp magnitudes.
Hi Ken et al.,
Funny, I (AHM) just wanted to shoot an e-mail myself when yours popped up...
According to Gary's suggestion ("mini-campaign") I am using the ensemble 80, 98, 105, 121, 122 and 125 with AIP4Win since this is the whole available sequence for that field of view (I don't have the chart no. handy right now), where AIP4Win is choosing the 125 star as check star (I thought, though, that the check star should not be part of the ensemble, but this is what AIP4Win supplies...).
I was afraid that 80 could show some saturation, but I tried to use just the last 4 stars with no apparent difference.
Some of you are using just one comp star. What are the thoughts about using the same ensemble?
My opinion is that we MUST use the same comp(s). I also think we should use an ensemble BUT some observers may not be set up to do that with their photometry software. In that case, one comp is necessary. This is not the best way to ensure accuracy but it certainly provides the best consistency among results from all observers.
It would be nice for all observers to report IF they can only use one comp during reduction. IF this is the case, the group could select one preferred comp close to the target and about the same color. Ultimately. we will need another one when the target gets fainter. We could limit the number of comp changes to make comparison easy.
So, can current observers report IF they can only use one comp? Perhaps we can wait a few days (5 days?) for reporting and then make a recommendation as Gary did? Note that two of those comps yield target results that differ by 0.1 magnitude, so only all four shoud be used in an ensemble OR the result agreement will be poor as we are seeing!
In the mean time, Gary's recommended ensemble may be best choice?
Ensembles can be useful for reducing random noise. The error will decrease with the Sqrt(N) such that if we go from 1 comp star to 4 comp stars we cut the error (uncertainty) in about ½ of that contributed by random noise (which is not the only source of uncertainty, by the way); Ensembles are also advantageous where we can average comps that have relatively poor photometry; poor color matches or values several steps removed from the target in that the problems created by these systematic issues should improve with the average function.
In the case of this specific target, using a bright comp (V = ~8) with fainter comps (~ V = 12.5) by an observer requires such short exposure to keep from saturation the V = ~8 comp that the observer is creating a great deal of noise with the fainter comps simply because of the shorter exposure. IMO, benefits from using this specific ensemble could very well be compromised in this specific situation.
I would also challenge the statement ~ “that one comp is not the best way to insure accuracy;” It may or may not be depending upon how accurate the original calibration was and how stable the single star is and how close is it in magnitude and color to the target.
The same can be said of ensembles in that they are not necessarily the best way to insure accuracy for many of the same variables in the previous paragraph. Often ensembles simply provide a different answer than carefully accomplished photometry with a single, well calibrated, stable, and well chosen single comp star and vise versa.
While we can compare the uncertainties I am skeptical that any one could point to one result as opposed to another and declare which is the most accurate. At best we can only compare the herd with those that are apart and wonder why the separation.
In looking at this specific Nova it is not to difficult, within some time periods, to spot those that are apart from the herd; in my experience the majority of these differences can be, for the most part, accounted for by any one of 15 (or more) individual choices, equipment combinations or possibly a mistake made by an individual observer; ensembles vs a single comp star is only one of those 15 items with the most frequent being saturation followed by under sampling to account for the differences.
Here are two random samples (you can probably find others that will refute this or support it, in either case but they do portray what I am trying to convey):
Here is what I mean by the herd, you will note that they are quite close in time and close in values.
Four observers back to back in time: 595.60337-595.60355
1) 11.158 Ensemble
2) 11.147 Single Comp
3) 11.161 Single Comp
4) 11.134 Ensemble
Here is an example of what might be suspected as an observer being apart from the herd:
Three observers back to back in time: 595.47023-595.470984
1) 11.189 Ensemble
2) 11.151 Ensemble
3) 11.070 Ensemble
As I mentioned you can find many other examples, some obvious and some less so and sometimes there maybe be multiple observers not with the herd. Point being, that it is not rocket science to spot observers who, for whatever reason, are marching to a different drummer and it is quite an open question as to what influence a single comp vs an ensemble might play in these situations; I suspect that, again, there could be any one of over a dozen possibilities.
[It should be pointed, out, that we must keep in mind that the target, after all, is a variable, and that, in itself, can sometimes account for differences]
I do agree, that in theory, if we could have all observers use the same ensemble, or the same comp star, that it would make, in general, for easier alignments…. Except that you will still have observers, for one or more of a variety of reasons still not agreeing with the majority.
IMO, it is a misplaced belief to presume that ensembles automatically trump, in a manner of speaking, single comp star photometry. In either situation it is really going to depend on the choices and care made by each individual observer.
As to which is more accurate, if in fact that is the right term to use, in all situations I would be interested in reading any published study or paper that compares both methods (ensemble vs single comp star), under a variety of circumstances, if anyone can point me in that directon.
Tim Crawford, CTX
I've seen your list of factors which influence magnitude calculations before. They are all valid to different extents. In this case, I do not think they are the cause of the observed variability of the three mentioned observers which included myself (MZK). We have been getting tight agreement until these recent measurements which seem to occur at a time when observers are selecting new fainter comps than during the beginning brighter phase when 80 and 98 were the only choices. Now many more comps in the 10-12 mag range are coming into play. I have seen that some of these comps appear to have a greater error than others (121 and 122 are particularly problematic).
IMHO, it is apparent that the greatest source of error at this point in the light curve is the selection of comp(s), be that one comp or an ensemble. Assuming the other factors are under control (as I think they are for this group of observers), the difference in comp selection is probably the source of the observed difference in magnitudes.
In your email, you seemed to agree that an ensemble (average magnitude from many comps) may improve accuracy (agreement with the "true" magnitude) if care is taken with selection with respect to magnitude and color. As a chemist, I think of this as a multi-point calibration curve analysis with interpolation of the target value. It is the only way I analyze samples. I never use a one-point calibration, except under duress!. Yes, I am comparing chemical analysis to photometry but I think the same criteria apply to all scientific measurements.
Obviously, if one selects one comp (standard) that has been "perfectly" characterized, the accuracy will likely be "best". BUT, we generally do not have any supporting information to make that selection.
Therefore, I continue to believe the use of a carefully selected ensemble is most likely to give a "mean" magnitude for the target which, in fact, best represents the most accurate value (i.e., closest to the "true" value). More importantly, it certainly will give a more consistent set of data since any inherent differences in individual comp magnitudes or colors will be treated identically when everyone uses the same comps.
I believe that most well designed campaigns by amateurs or professionals select the comps to reduce or remove extraneous/unknown sources of error. Yes, I do ASSUME that most egregious hardware errors are not present in the data of experienced observers.
The sequence team just uploaded a new sequence that goes fainter for V339 Del. There were also some revisions with this update to the exisitng 12+ comps.
Tim Crawford, CTX
Thanks for the new sequence!
There appears a double 122 in the table (12841ES), but only one 122 in the chart; and I cannot find 143 in the chart (12841FN). Should that be corrected to avoid confusion? Or am I the only one to get confused .
My charts are and have been printing correctly. I also just re-accessed our VSD data base and confirmed that I had originally taken the following two actions:
1) The 000-BLC-829 122 comp had the label suspended as part of the new sequence upload for two reasons: a) the new data did not include that specific star and b) it was a duplicate.
2) The 000-BLF-179? 143 comp was part of the original sequence upload but when I immediately looked at a printed chart it appeared to be two close of a double so I completely deleted it and replaced it with a the 142 comp, although this took me some time (10-15 minutes) to accomplish given the nature of the data and that crowded fov.
I can only assume, always a risky proposition, that you managed to make some chart downloads in between my changes and caught a chart before the duplicate was removed as well as a different chart before I could replace the 143. Just a question of bad timing, IMO.
Please try making new charts again. If you still have a duplicate 122 star then email me privately and we will get to the bottom of this; obviously you are not going to find the 143 star.
To all observers: while this is a "live" topic, and I can appreicate an observer's desire to prevent an error, we have routines in place to deal with suspected discreptant sequences and I encourage everyone to use that CHET tool so that forum topics can avoid going sideways, as sometimes happens (although I do not beleive it occured here):
Per Ardua Ad Astra
Tim Crawford, CTX
Hi NovaDel2013 analysts, NovaDel2013 interests me a lot. I discovered very soon that Arne Henden suspected some variations at the time scale of the hour. Me too. I contributed to AAVSO visual observations only, as far as I could it (observer PREB). As an aperiodic star, NovaDel2013 needed a new tool for visual analysis. Especially devoted to the detection of eventual pulsations. I wrote it and called it "Intraday". I let it on "http://etoilesvariables.fr/index.php" (today on its start page, later under "Materiels et logiciels"). You are free to use it, to pick and adapt freely its source code to your convenience. The main capabilities are: _ resize of the window, _ animation (scrolling of the image at a variable speed), _ gradual connection or disconnection of a 'curve'. Of course, Intraday: _ manages the colors, _ zooms manually (like Zapper) or automatically (centered zoom), _ reduces the image (reciprocal of centered zoom), _ translates the image with the mouse, _ filters the bandpass (vis,R,V,B,I,...), _ is compliant with the AAVSO download files, _ compares different observers or groups of observers, _ may be used with other stars. The main goal has been to detect some intraday variations, as soon as possible: juxtapose 2 runs of Intraday, one under the other, with the same zoom ratio and the same scrolling speed. But at 2 different dates, in order to discover a curve shape similarity. I used Intraday myself, but unfortunately, I cannot answer the question of eventual periodic pulsations. Sorry. I tried to protect it against bugs, but who knows... Intraday has 3 prerequisites: _ Microsoft Windows API, _ Open GL, _ C++ language. It works on Windows 64bits. A different build works for Windows 32bits too. You may find in "ZipNova0.zip": _ a ReadmeE.txt file in english; _ the source code, together with the 2 .bat files for the build; _ the binary executable "Intraday.exe" for Windows 64bits. After renaming of the data file "NovaDel2013_1_Full" into "NovaDel2013_1", you may run: _ either "Intraday filter=vis" from a command prompt, _ either "Intraday filter=B"... You may add a "NovaDel2013_2" file as a subset of the whole, for instance your own observations. I'm happy to share my research with you. Pierre
Attached is VPhot photometry page which shows variation in mag for each comp in time series compared to quoted mag. All agree within 0.03 mag except the 120 comp which differs by 0.09 mag. It probably needs some tweaking! Probably not the best comp to use if only using one.