We had good success on the Nova Forum in critiquing observers and improving the photometry. Unfortunately, Nova Del is in its nebular phase, which means almost all of the flux from the star is through emission lines. These make comparison between observers almost impossible, as you can't transform cleanly and your particular filter/camera/telescope combination may have a different throughput for the various emission lines than is the case for the next observer. Continue observing it (please!), but perhaps it would be good to choose another star for critiquing observers.
So I'm going to have a few test stars over the next year, and I offer my expertise in critiquing your observations. These test stars will be non-pathological: no emission lines, no strong molecular absorption lines, no flickering. This limits us to primarily regular pulsators and eclipsing binary stars. However, the goal is for everyone to submit observations that fall within 0.01-0.02mag of everyone else. It can be done!
Rule number 1: you have to make your observations as "sets", with a minimum of 3 images per filter. Results need to be averaged and the report needs to be the average value and the standard deviation from that value. I'll post Richard Sabo's Windows program on the website as a helpful tool.
Rule number 2: you can do single star or ensemble, as the comparison stars for these field have well-determined magnitudes and colors.
Rule number 3: you need to observe in at least two standard filters. I'd choose B and V as the best; V and Ic as second best; any number of filters is ok. I'll probably do BVRI myself, but may include Johnson U, for example.
Rule number 4: you need to transform your data. This is the hardest part of the experiment. You can post untransformed data at the beginning, but learn how to transform, get your coefficients, and replace the untransformed data with properly transformed data before long.
The first object: XZ Cet. This is classified as an anomalous cepheid, with a period of 0.823 days. At 2hrs RA, it is an early evening object. You do not need to have a cadence of more than 10-15 minutes, and once/hour or evena couple of times per night is fine. So you can go off doing other projects inbetween your measures. I've posted a comparison star sequence. None of the stars is comparable in brightness to the target, but you can adjust your exposures so that the target is not saturated, yet you have good signal/noise in at least one comp. Note that most of the comps are to the east/northeast of the variable, so you might want to offset the variable in the field to pick up comparison stars. The variable is 9th magnitude, so watch out for scintillation; it is at -16 Dec, so don't observe past 2-3 airmasses (that in itself may limit how long you follow it!).
Good luck! As observations are submitted, I'll post on this thread any comments that I might have. I expect others to contribute their views as well! If you don't like this star, wait for a month or two and I'll have another one for you.
Arne
Arne:
Can you suggest a methodology for PEP observers who would like to participate? It would be interesting to see how PEP compares with CCD (and each other). Our cadence is intrinsically much longer, and we may have access to only a V filter.
A suggested pair of comparison stars to use as a standard would be welcome.
Tom Calderwood
Nice
To: Dr. Arne,
Nice experiment! I'll be glad to participate if I can use the software package LesvePhotometry to perform the transforms on XZ Ceti. I'm using Maxim DL 5.23 to reduce my data. Is there any other recommended (non-cloud) software package I can use to reduce/get transfomed BVRI magnitudes?
James Foster, Los Angeles, CA
Hi Arne,
Thanks for choosing a target accessible from both northern and southern hemispheres, and bright enough for DSLR observers. I will participate in this project and I'd like to encourage other DSLR users too.
Could you include additional, relatively bright, comp stars out to say 2 degree radius from XZ Cet?
Cheers,
Mark
Hi Mark,
I'll see what I can do. The field was imaged primarily with W30, with a 20x20arcmin field of view. I'll put it in one of the Bright Star Monitor programs to both get a wider field for the DSLR folks, as well as pinning down a couple of decent comparison stars for the PEP observers. That may take a week or so.
Arne
Hi Arne,
I submitted a short series of DSLR observations a couple of days ago. They're only preliminary at this stage, I'll recalculate derived magnitudes if/when more bright comps are added as we discussed earlier.
For the submitted data I used the 98 star as the Check and an ensemble for the Comparison including the 99 star plus another dozen or so stars brighter than 10 V mag within about 2 degrees of XZ Cet. These were selected using SeqPlot and checked for variability by refering to SIMBAD. The Tycho 2 values from SeqPlot were used for B and V, but R magnitudes had to be estimated based on the B-V color.
Images were recorded a couple of weeks ago near full moon and transparency varied throughout the original 4 hour time series. I opted to use just an hour of that data where things were relatively stable. Each exposure was 120 sec at f3.2 and ISO 400 through a 200mm lens. The reported mags are averages from three images and error is the standard deviation of these three measurements. All are transformed magnitudes, but extinction correction was not applied.
Sydney weather has been uncooperative but I'll image XZ Cet again when I can. Cheers,
Mark
Dear Arne et al
I am happy to contribute to this project as the star is convenient for southern observers as well.
I have regularly transformed my data using a single comp star and understand how to do this when I measure raw magnitudes with either AIP4WIN or MIRA.
I have started using VPhot and it creates ensenble measurements. It can also transform and I have been using this with nova del. This is fine for occasional measurements but uploading large numbers of files is difficult and time consuming with a slow internet connection. I would like to make these transforms in a spreadsheet like I do with a single comp star.
I'm not sure how it transforms using an ensemble as there isn't a real b-v measurement for an ensemble.
If I wanted to use an ensemble with AIP4WIN or MIRA and a spreadsheet, what is the method to do it?
Cheers
Terry
Arne,
Is there any problem using NGC 7790 as a standard field for calculating transformations? It is well positioned at the moment. Or, since the new Landolt +50 standards are out, would it be better to use a selection of those over a fairly wide range of B-V values with suitable extinction corrections? Berry and Burnell not withstanding, I suspect the former for those of us with, at best, marginally photometric skies.
Brad Walter
Arne:
I'm quite excited about this project and looking forward to submitting soonest. I will probably try several different processing pipelines and compare results.
You mentioned that you would provide a chart - is this a special one or should we just use the photometry table from VSP (mostly the stars around BD-16 351)?
Rick Wagner
Hi Richard,
The comp stars are already in VSP, so they are primarily the ones around BD-16 351. There should be plenty to choose from, if you can get them into your field of view.
For those of you who do not have a "transformation pipeline" set up, I'd recommend submitting your untransformed data for now. The field is heading west, and the sooner you start, the better. You can determine coefficients and learn how to transform later, and then delete/resubmit the improved photometry. In the meantime, you collect data, and correct some of the basic problems that don't pertain to transformation anyway.
Arne
Arne,
I was about to start participating in this, but VSP shows no comp stars (on chart or on list of comps) in a 20 minute field. Am I missing something?
(Also its a bit low for us mid northerners)
Dave LDJ
[quote=ldj]
Arne,
I was about to start participating in this, but VSP shows no comp stars (on chart or on list of comps) in a 20 minute field. Am I missing something?
(Also its a bit low for us mid northerners)
Dave LDJ
[/quote]
I also found this but using a 1deg field they become visible. I took images last night of the field. My image size is about 20x30 arcmin and I was able to have the variable and 3 comp stars in the field with the stars a reasonable distane from the edge of the frame.
I assume the choice of a low star for you northerners was deliberate as it will exaggerate the colour difference with comp stars. As the purpose of the exercise is to improve accuracy between observers this will highlight errors. There are less observers in the south.
Terry B
BHQ Armidale NSW
Dave, as I mentioned earlier, you will have to offset from the variable. If you enter XZ Cet on the "Star Finder" part of the home page, you get a 1-degree field. That shows you where the comps are; mostly to the E/ENE. You will still be able to image the variable, just off-center.
Arne
As it turns out, two
Arne,
As it turns out, two of the comp stars just fit diagonally in my C14/ST8/compressed field, assuming the pointing works well enough. I'll start observing it next clear night and submit non-transformed data with the goal of using the project to learn how to do transformations as I have not done that yet.
-- Dave LDJ
Hi Terry,
There were several reasons for choosing this target; one was to both give southern observers something to image while at the same time giving northern observers something more difficult to measure properly. The lack of nearby comp stars is actually an interesting aspect. However, to me, the primary reason is that this is a really neat star that has garnered zero attention from the AAVSO observers. I have hundreds of images of the cepheid and would like to publish a paper on it, incorporating as much new quality photometry as possible.
Arne
It is interesting to compare XZ Cet's lightcurve to those of the Anomalous Cepheids in the LMC obtained in arxiv.org/abs/0811.3636 (Figure 4). XZ Cet would appear to be a first-overtone Anomalous Cepheid by their classification.
Cheers,
Doug
As it happens I'd taken some observations of XZ Cet on 5 November, so I've started early. :-) I've reduced these again using the comp stars Arne has identified, stacked the images and uploaded the results. XZ Cet was central to the image, so the C and K stars were a bit too close to the edge of the frame for comfort -in particular the check star was in a zone where vignetting was beginning to be an issue even with flats.
Forewarned now, I'll centre the image to the NE of the target.
Arne:
Ran two BVRI sets (3 per filter) of images last night separated by about 30 minutes. I took them on either side of the meridian and thus was able to keep them at about 30 altitude.
The following observations are made:
I offset image FOV to the NE to get six comps and target on about 30' FOV.
Three of six comps do not have R or I magnitudes.
I ran ensemble using all comps (- 1 as check) for B;V. However, R;I magnitudes only had one comp due to lack of reported magnitudes.
Stacked the 3 images for each filter with VPhot.
Ran color pairs to transform data in VPhot. Thus, I got three V magnitudes from BV, VR and VI pairs. Selected only one V to report. Total difference range in three V magnitudes was about 0.003 for one set and 0.008 for other set.
In 30 minutes, I could see that the target got brighter by about 0.04 mag, which is more than error.
Hello Ken
Did you mean fainter in your last sentence? Looks like it from the data on 2456613 on LCG. Maybe I don't understand your comment? Did you take out the extinction and transform?
By the way, excellent agreement on that night.
Gary
Oops! Yes, it was fainter. I forgot which image set was first. ;-(
Hello Arne and All
Just uploaded a second night of observations of XZ Cet (JD 6616). This time with the CDK17 at MMO. The B and V filters are the same Gen 3 as my observations from JD 6613 a few nights ago. The data last night is cleaner, although the night was poorer. Both nights were binned x3 or x4. There is some offset.
One can clearly see either or both the anomalous cepheid behavior and/or the extinction.
Gary
WGR
PS: Just added the second half of last night. It was on and off clouds, but there were ok patches. Data looks about the same, but noiser. Trend about the same.
Here is the V-band light curve output from VStar, using the VSX period. As you can see, we're starting to fill in the light curve. I'll be uploading the single-visit BVRI data from BSM_Berry in a day or two so that we can see how that telescope works.
Arne
Hello Arne
Is there a way to get Vstar to plot the diagrams of B or I data? I guess one could download a file and do it that way. Its just so convenient to do the V data. Something to add for the future.
Gary
Looks like I got the pre-peak, peak, and post-peak of this anamalous cepheid last night. Got it in 3 colors. The red seems to differ from B and V. The beginning of the peak is there, but the tail end is different.
Also looked on VStar and plotted the phase diagram with the defaults from VSX, and we are starting to see the variation in periods that was mentioned in the literature.
Gary
WGR
Hi Gary,
I'm not a VStar expert, but when you download the dataset, there is a clickbox at the bottom that says "all data". If you click that, then all bands get downloaded. Then you can go to View and Filter and enter band, equal to, Johnson B, and you will see the B-band measures colored blue and the V-band measures colored green. Then you can phaseplot and see both.
That said, I notice one dataset is considerably offset than others. I'll have to look at this soon. I've been waiting until we get the multi-night dataset from BSM Berry processed so that I can compare those transformed values with other observers. Soon!
Arne
Hello Arne
I assume that this conclusion is based on looking at a phase plot. Since the variable has a range of at least half a magnitude according to the lit, it would be hard to determine this from the LCG. I did also notice from the Lit that it undergoes period changes and other behavior. How do we separate out these variations from observer offset? Perhaps taking only one observer at a time and see if the phase plot is stable. Looks like "needs more observations?"
How large is "Considerably offset", much larger than .01-.02 I would guess. (I don't have Vstar/Java on this computer--I am on the ferry, returning to America from Nantucket.
BTW: I have included observations from 2 sites--I wonder if this is causing problems?
Gary
Hi, Gary,
>>> I did also notice from the Lit that it undergoes period changes and other behavior. How do we separate out these variations from observer offset?
The period changes, phase shifts, light curve changes, usually affect the time axis not the mean magnitude so magnitude offsets so any meaningful difference in mean magnitude will very likely be instrumental.
This is the ASAS-3 light curve between 2000 and 2009 with a period slightly shorter (0.823140 d.) than that in VSX (from combining Szabados and ASAS-3 data). As you see the changes are not so dramatic so unless it is now changing its behaviour, I think photometric problems can surely be noticed.
Cheers,
Sebastian
Hello Sebastian, Arne and all;
I was thinking when the period changes, there is a shift to the right/left of the phase diagram, which can look like an amplitude change/shift if the data is not complete. After your comment, I took another look, and I agree with you that this is an amplitude shift for sure. It looks to be about 50 millimag PT.
The data is taken by one observer on 2 ajoining nights--same equipment--observer is me! Then it dawned on me what might be going on here. The first night I used the standard convention, left-right is ra and up/down is dec as I almost always do. For my setup, this puts XZ Cet and the 107 Comp very close to the edge of the chip. On the second night, took twilight flats as normal (JD 6617) and started to observe. Decided to rotate the camera about 20 degrees so that the stars would be along the diagonal of the chip, and as far from the edge as possible. Then reduced the data with the flats from that night (which were not rotated). I need to reduce this data with the 20 degree rotated flats, and I am hoping that the offset will be eliminated. I won't be back at that telescope until 6 Dec, so I will fix that data at that time. In the meantime, back in America, awaiting another clear night.
Doug, your suggestion to use Vstar helped a lot here. Wa able to identify the observer as myself and see what was going on.
Gary
WGR
Hi Gary, Arne, Doug, all
I'm happy that VStar is of benefit to the XZ Cet experiment; it looks quite interesting by the way. I must read all posts to this topic in detail.
If anyone has questions about VStar, don't hesitate to ask if you are uncertain.
Hopefully the recently completed user manual will also help:
http://www.aavso.org/files/vstar/VStarUserManual.pdf
Sections of particular interest here might be:
Thanks and enjoy!
David
I just want to emphasize how great VStar is for looking at these datasets (compared to LCG). The learning curve for VStar is quite short and one can select a point in a phase plot and immediately see all of the details about it in the Observations pane.
Cheers,
Doug
I've made four
Hi all,
I've made four sets of observations so far (5, 18, 18, 23 Nov - the 5th being a coincidental observation I made before this started). Frustratingly, because of cloud and some equipment issues, I have only been able to get a short set of observations in each case - 12-20 20-40s images in each of B and V, which I've stacked in sets if 3-4.
Given that even these short runs are enough to show brightening or fading of the target, I assume the stabdard deviation we are to report is in the check star. But over what time period? I'm happy with the precision I seem to be getting (even thoguh untransformed, my obs are sitting nicely in the phase plot) - 0.001-0.004 mag error as reported by MaximDL based on SNR, and standard deviations on the check star of 0.003-0.008 mag. If tonight as planned I manage to get a good 2 hrs on the target, should I report the standard deviation of all those observations, or of subsets?
Lastly, is there an idiots' guide to transformation for differential photometry out there somewhere?
Thanks again for the project commentary. As a beginner to photometry I have learned an enormous amount folowing this and the Nova Del thread.
Cheers
Jonathan
Hello Johnathan
Glad to see you active on this object. It does not have the draw of Nova Del.
The problem I see if using "standard deviations on the check star" in this case is as follows. I cannot get a check star of comperable magnitude to the variable. I am using the 107 as the comp, my check is 12-14th, and so its poission reported error is much larger than the poisson error of the variable. It basically contaminates an otherwise very good observation. It does give an upper limit, as my errors are probably much smaller. Of course, the Poisson errors are not the only ones, nor the dominant ones. These other errors are of course, mixed in with the observations.
Some folks use the poisson error of the target and add it quadratically with the C-K value. Again, if the C and K is much fainter than the variable, its overly pessimistic. Remember, that since this is differential photometry, the software measures the Comp, and the Target, calculates the difference, and adds that to the Standard Value of the Comp. So the errors in the Comp are included in the measurement of the Target, by definition.
I personally like taking several measurements, 3-10 typically, and take the standard deviation of the mean. This assumes that the star is changing brightness slowly enough that its not contributing, and we are getting the poisson, the seeing, the a/d, the software, etc. This can be done with a tool that Bins the observations. I know Arne is working to get one on the web site. How to calculate the errors is part of the issue.
Gary
If the star is varying too quickly to get a standard deviation from the mean of several observations (as is the case at least part of the cycle for almost all of the variables I follow) perhaps fitting a straight line to a half-dozen or so observations and then measuring the scatter about the line would give a better estimate.
Rick Wagner
Thi would seem to be a reasonalble approach for this special case.
Gary
This is in reply to several earlier posts. I've been busy working on other projects, and have not chimed in recently. Sorry!
The phase plot is now just about completely filled in. I attach the latest Vstar version to this post. It is pretty obvious that either (a) the star is changing its amplitude constantly, or (b) we're not there with transformation! The other observation is that there are a couple of people with very large error bars, for this 9th magnitude object.
The people I see that are transforming: SRIC, CTOA, MZK
The people I see that are NOT transforming: LDJ, WGR, PJOC, AHM, BHQ
Three people have observed through minimum, with values
SRIC 9.681
LDJ 9.817
PJOC 9.756
overlap measurements: AHM and WGR are about the same; WGR is 0.05 fainter than LDJ; LDJ is 0.13 fainter than SRIC(t). MZK(t) is 0.08 fainter than WGR. Therefore, there is an inconsistency between SRIC and MZK, though I'd like some more overlap. PJOC and BHQ don't have a consistent offset with each other.
So you can see that there are considerable offsets between observers. Each observer by themself looks to be consistent.
Jonathan asks:
[quote=PJOC]
Given that even these short runs are enough to show brightening or fading of the target, I assume the stabdard deviation we are to report is in the check star. But over what time period? ... If tonight as planned I manage to get a good 2 hrs on the target, should I report the standard deviation of all those observations, or of subsets?
Lastly, is there an idiots' guide to transformation for differential photometry out there somewhere?
[/quote]
In general, check star standard deviation is used for reporting, as it is assumed to be a constant star measured in the same manner as the target. For a short time series at low airmass, you can use the standard deviation from the average value of the (C-K) difference. For longer time series, there are complications; the main one is changing airmass, where the scatter will typically increase as you head towards the horizon. You may also have clouds during part of the time series, but clear the rest of the time. In those cases, you can be selective, and break your time series into a set of "windows", where within each you report the standard deviation from the average value of (C-K). This is "ad-hoc"; the main concept is to find a method that best measures the uncertainty of the observation.
As Gary mentions, the C and K stars may be fainter than the variable, and therefore their standard deviation would seemingly overestimate the uncertainty in the target measurement. In practice, I don't find this to be the case. Usually the other error sources besides Poisson dominate the uncertainty, things like centroiding, changing seeing, wisps of cirrus, crossing hot pixels, cosmic rays, etc. My guideline is that it is always better to overestimate the uncertainty than to do the opposite.
I also like Richard's suggestion of fitting a line through several target estimates and looking for the standard deviation from that line. That presupposes that you are in a linearly-varying part of the light curve, and not, for example, near max or min.
Regarding what cadence to report your observations, the usual guideline is to choose that cadence that shows variation. For a 0.82day period object, if you chose 1/100 period as your measuring cadence, that would be one observation every 12 minutes or so. If you look at the light curve, you can see that there are phases in which the variation is more rapid than other phases; for example, this is an assymmetric light curve with faster rise than fall. You could sample at 1-minute intervals and have noticable variation between the samples. Therefore, almost all sample rates are "ok", with one big caveat: if you have high cadence, you should also have high signal/noise in each point. This is not a flickering system; you want high quality datapoints, not a cloud of datapoints. So if your photometric errors are 0.05mag per point, I'd suggest averaging several points together, even in a time series, and report with lower cadence.
As for an idiot's guide to differential transformation, there isn't a whole lot available right now. Let me see if I can get my book chapter cleaned up enough to post to give you guidance. I also have some old XZ Cet photometry from my prior life that I'd like to show, and perhaps use this as the "comparison set" to get people on the same system. But, first on my list is uploading the best calibration set for NGC7790 (for the northern folk).
Arne
Arne
Just a correction. My measurements (BHQ) are all transformed. I did submit 2 sets of data that were transformed but labelled in the database as not transformed. My mistake when submitting them. I have corected the database.
Cheers
Terry
Arne,
In your initial post you mentioned that you were going to post Richard's program as a helpful tool? What is the program for and can you post a link for it?
Barb
Hi Arne,
last night was clear so another time series was recorded, reduced and uploaded. The comments in my previous post apply to this dataset as well. Once again transparency was variable so extinction correction not done.
Hopefully tonight will be more stable. Cheers,
Mark
I've uploaded observations from the first Sonoita Research Observatory, a 35cm system with an STL-1001E camera (the same camera being used today with W30). This is just the BV piece; I'm still refining the R,I comparison star magnitudes. Approximately 2300 datapoints from 2005 are available. If you use VSTAR, limit the data to between 2453609 and 2453673. This will give you an idea of the results from a single telescope. I'll start describing the differences between the current observations and this baseline light curve in a day or so.
Arne
To start the discussion, I've attached the SRO35 phased light curve as well as the recent AID phased light curve (using VSTAR) to this post. Let's look at the SRO35 light curve, which contains about 1100 measures over a 63-day window in 2005. Peak is prior to the current ephemeris (and gets worse for the recent data), so probably the period is slightly in error. There is more scatter near maximum, which is probably real. I'd like to see the Rc light curve, as often there are shocks propagating through the atmosphere that lead to emission. For B and V, the peak is nearly coincident; there is often a color phase lag on pulsating stars.
Note that the amplitude in B is more than at V, which is typical of pulsators, and that the star is redder at minimum, again typical of pulsators where they are the hottest near maximum. The curve is asymmetric with faster rise than fall, and the peak is broadened with a shoulder after true maximum. I've deleted about a dozen data points from the original set, as they had too large of error (this field transits about airmass 1.7 in southern Arizona, so many of the observations are taken at quite high airmass and not always photometric weather). Typical scatter on this light curve is about 0.03mag peak-to-peak, which equates to about 0.01mag RMS deviation from a mean curve. Not bad for a high airmass field with few good comparisons! In this case, I'm using some recent calibrations from BSM_South for about 11 stars, ranging from V=10 to V=13 in the Sonoita field of view. That is cheating; I'll upload the calibration photometry soon so that you can take advantage of it as well.
Note also that this is an idealized case - all of the observations were taken with the same equipment and properly transformed to remove any systematic effects like airmass. If you look at my original data, you will also see a bug in my differential photometry code - all measures are reported at airmass 1.0. I'll try to fix that.
Now lets look at the AID curve. There are two obvious features. Certainly there is an offset between observers; in B, the difference is as large as 0.15mag. V tends to be a little tighter, though even there a couple of observers are off of the mean curve. The other main feature is that many of the observers are reporting quite large error bars; 0.02-0.03mag is shown for many of the points. For a 10th magnitude star, the errors should be very small, so we need to look at those observations and find out why - are the exposures too short, is the airmass too high, are they using an ensemble with large deviation between members of the group? Finally, as mentioned before, the period probably can use some adjustment.
I'll cover individual observers in the next post, probably tomorrow or Thursday.
Arne
Hi Arne,
does it make sense for all observers to strive for a similar FWHM to reduce scatter?
Gianluca
Hello Gianluca
THought I would try to give an answer to your question about similar FWHM to reduce the scatter. I think that as long as all observers use an aperture that has a radius of 2-3 FWHM, then the scatter introduce by the different FWHM's should not be large. This assumes that the images have sufficient exposure to have similar SNR.
Gary
There often isn't a lot you can do to make the FWHMs constant within an image or between images. I have an aluminum tube uncorrected F12 Dahl Kirkham telescope (Mewlon 250) and a rather large format camera 3000 x 2000 pixels at 9 microns/pixel. That gives me a plate scale of about 0.62 arcsec per pixel and FOV of about 30' wide by 20' high. That is large for an uncorrected DK, resulting in significant difference in FWHM over the Field of view. My location has pretty bad seeing most of the time, as well (3 to 4 arcsec FWHM). I can literally see the random deformation of star profiles across the image "blobbing out" every which way without an apparent pattern. Also the aluminum tube means the focus changes considerably with temperature. An IRVBBVRI image set takes about 10 minutes for the integration times I am using for a bright star such as XZ Cet. When the temperature is changing fast, as it tends to do between dark and about 11 p.m. on Texas winter nights, I should probably re-focus after every two or three image sets. Sometimes I do but, if I have a lot I am trying to get done I may not, and the FWHM creeps up. Alternately, after midnight the humidity may increase rapidly and the primary mirrorstart to collect dew which really affects the FWHM. It goes from ok to bad in a few minutes. If I look out of the telescope shed on those occasions, I can often see the fog starting to roll up the hill from the creek bottom on the far side of the pasture.
So, If you can't do anything about it. How much does it matter? Attached are some Excel spreadsheets of untransformed XZ Cet B data from 12/16 and 12/18 UT. The weather was clear but the seeing ranged from mediocre to poor, even for my area. Two of the spreadsheets contain FWHM data for XZ Cet, the 99 star and the two 107 stars. In the spreadsheets each row contains FWHM data for the 4 stars in a single image. Each FWHM column contains the FWHM data for a single star across the images. Averages standard deviations, min, max and max-min values were calculated for FWHM values for stars within individual images and individual stars across all images. Averages of all of these statistics were also calculated. For example, the average of standard deviations of star FWHM within images and average of standard deviations of individual star FWHM across images were calculated. The 11/16 FWHM spreadsheet is for a series of 7 images taken between about 20:20 and 20:45 local time. The seeing was quite poor as can be seen from the high average star FWHM within individual images and the variability of the FWHMs of individual stars between images and the variability of the average image FWHM between images. Things were jumping in a pretty lively manner. On top of that there is a pretty wide range of FWHM for stars within individual images which can be seen in the relatively large standard deviation values at the right hand end of the spreadsheet. If that isn't bad enough the sky background level was high because the nearly full moon had been up for about 2 hours. Yet relatively precise photometric results of 0.006mB could be achieved based on the standard deviation of measured values for the 000-BLF-0056 comp star (107 star nearest the 99 star). I submitted the values measured with an aperture radius of 17 pixels with the inner and outer edges of the background annulus set at radii of 25 and 40 pixels. The measurement aperture was not arbitrarily selected. If you look at the Growth Data spreadsheet for the 12/16 V data, I used the aperture that seemed to provide the best overall results. I use the following 3 criteria and there is something of a balancing act between them;
1. Growth of the target magnitude had steadied to a point at which additional changes were small in relation to the claimed uncertainty
2. Same for growth of the check star. Growth had essentially stopped and started reversing with an increase of two pixels in radius
3. The uncertainty in the measurement of the check and target stars at or very near the minimum (at the minimum in this case)
The growth data is extracted by making multiple measurement iterations. A set of measurements for XZ Cet, 99 (000-BLF-055) and 107 (000-BLF-056) for all 7 images with a single aperture is contained in spreadsheet Photometry_XZ_Cet_West_B_17-25-40-0.xlsx. It is interesting to note that, as is often the case, the Error value generated by the program using the full CCD equation underestimates the measured uncertainty of the check star measurements, in this case by almost 60%.
Three similar spreadsheets are also attached for V data taken two days later. In this case the data acquisition started at about the same time of the evening but continued for several hours including a break of about an hour after the first two measurements to flip over the meridian, re-acquire the star and re-focus. This FWHM data shows that the seeing was improving as time progressed and the variation in seeing was also less erratic. The same process was used to determine the best choice of apertures. Because of the break between images and the obvious difference in FWHM, I did separate analyses of the growth data before and after the flip. The growth data attached is only for the 8 images taken after the flip. The choice of aperture is a toss-up for radii between 11 and 13 inclusive. I picked the large aperture end of the range. Someone else might pick a different one. There isn’t much difference in this range. The resulting uncertainty determined empirically from the check star was about 0.005mV, a very small improvement over the B data two nights before even though the seeing and FWHM (for whatever reason) were better.. Once again the error value reported by the software underestimates the uncertainty in the check star data, but this time by over 70%.
So this extremely long winded explanation demonstrates, fortunately for us, that keeping FWHM constant isn’t all that important. Further there is significant leeway in picking the measurement aperture radius, but it is best to do the analysis to see how much leeway you have and in which aperture range it resides. Then you can report your results with much more confidence, but in my case, at the cost of several more hours of work since I don’t have all of this stuff automated in scripts and macros. I keep promising myself I will get around to doing that. However, messing around with the data provides invaluable insights into what matters and what doesn’t and methods of revealing those differences. I keep reminding myself of that as I iterate measurements and sort, slice, and dice spreadsheets.
Brad Walter, WBY
On 12/23 I took another set of data with much better seeing and no moon. The comp star FWHM was smaller than the target or check by between one-third and two-thirds of a pixel respectively but the full profile width of the comp above the background noise is in between the target and check stars.
The growth data characteristic is quite different. than for the two previous nights. Maximum magnitudes (max flux) were obtained at measurement aperture radii of only around 1.3 FWHM - unusually small for my area. However, uncertainty of the measurement keeps decreasing over a wide range of apertures larger than those yielding max flux for the target and check stars. Normally if all the uncertainties are decreasing all stars are picking up net flux relative to noise, so the magnitudes keep increasing. Usually the uncertainty starts to increase before the flux capture "rolls over" and starts to decline. One would then think that the increasing aperture is picking up proportionally more comp star flux than either the target or check star but the target and comp are about equal magnitudes and the width of the comp star profile above the background is half way between XZ Cet and the Check star. The check star profile is least skewed, XZ Cet is the most skewed and the comp is again in the middle by visual inspection of the profile plots. So this is one of those situations where you can't determine what is causing the unusual pattern in the growth curve but inspection of the images reinforces that choices near the points of max flux capture for target and comp should be the best choices. I applied the same criteria as for the first two plots and chose the measurements taken with an 8 pixel aperture radius. I think you could equally have chosen an aperture one pixel smaller or larger but the difference isn't significant with respect to the uncertainty of the measurements.
In this case I think the choice of max flux is clearly the better choice than lowest noise because there is a significant increase in magnitude (decrease in flux) before the minimum uncertainty levels are reached for either the target or comp.
Anyone else have an opinion?
hallo,
My name is Aniruddh Deshpande and observer cade is DAND. I have 10" aperture with focal length 1200. I am doing photometry using dslr camera cannon 7D. Getting 1 degree field.
Mr. Rojar Pieri is my mentor and he advice mi to participate in this experiment.
I donot use any filter.
That is unfiltered with V-zero point.
I need to check my accuracy of photometry, and what should I observe ?
How can I do this. Please guide.
Start here:
http://www.citizensky.org/content/dslr-documentation-and-reduction
I am just curious as to why VSP has no Rc or Ic sequence values for 99 (000-BLF-055) & 107 (000-BLF-056). There are values in SeqPlot although they have larger errors than values for the other 107 star (000-BLF-059) but no larger than the B-V error for the 99 star.
I am also curious as to why the uncertainties for the 99 star in SeqPlot are significantly larger than posted in VSP. Is this just a matter of update timing and frequencies of observation or is there some underlying reason for not using these two stars in Rc and Ic?
Brad Walter
Brad,
The magnitudes and colors for those comp stars come from BSM New Mexico. There is no Ic or Rc data. If you are clicking on stars in SeqPlot and getting R and I you probably have Tycho or some other photometry DB checked on. APASS won't help you because it saturates at those magnitudes and you're at the edge of reliability for Tycho.
From your site I wouldn't bother doing Rc anyway. Stick to BV and I.
Mike
Ah, perceptive as always. You are correct I downloaded APASS values and I didn't stop to think that the 99 star would be too bright. I thought APASS values might be OK down to 10.5 for 000-BLF-056 and 059 stars but I see they are not. What is the lower limit preference for APASS? It must be higher than the recommened minimum for SRO data.
SeqPlot doesn't have a B-Verr listed and I notices that the Berr in the sequence is the same as the Verr in SeqPlot. How does that work?
One final question. Why do you suggest not bothering with R values in my location? Since you mention my location, there is more to it than simply the high errors associated with R for 000-BLF-059 (107). I can see that the larger frequency spread between V and I has the potential for observing color changes and timing differences of maxima and minima. Does it have something to do with my low elevation?
Brad Walter, WBY
Hi Brad,
One reason people don't bother much with R anymore is that the amount of information ina BVI set is pretty much identical to the amount in a BVRI set. The extra color index adds essentially no additional useful information.
Cheers,
Doug
Thanks, Doug. Well, I thought that might be one of the reasons, but Mike referenced "my location", specifically. I was wondering what about my location might make a difference and the only thin I could think of was the altitude of about 130 meters above sea level. I am abouttwo hundred miles from the coast but still pretty low elevation. The Texas hill country doesn't start until about 40 miles NE of me.