Affiliation
American Association of Variable Star Observers (AAVSO)
Thu, 08/20/2015 - 16:50
AAVSO Alert Notice 526 announces a campaign beginning today (August 20) on the pulsating helium and zirconium-rich hot subdwarf LS IV -14 116 (V=13.0) in support of upcoming satellite and ground-based spectroscopy observations. Please see the Alert Notice for details and instructions.
Many thanks and good observing,
Elizabeth Waagen, AAVSO HQ
Great to see early data for this campaign. OYE's data look very good ... the amplitudes are indeed small -- but real. Other observers ... please watch your counts, and if you think you need more, then drop the V filter and concentrate on B. I have just arrived at ESO-Paranal and start observing tomorrow (23 August). Meanwhile, if any observers would like feedback, please ask (csj@arm.ac.uk).
The observed variability of this object is miniscule.
Did anyone look at the Jeffery paper referenced in VSX or the Kilkenny & Pauls 1990 paper? The amplitudes of the two observed modes of pulsation are about 2.1 millimags with a period of about 33 minutes and 1.9 millimags at periods of about 48 minutes.You really need SNR of 1000 or higher. With our typical pixel sizes (say 9 microns) and our typical seeing (say 2.5 arcsec) 10,000 net counts per frame for a star is going to give a pretty poor signal to noise ratio. In a campaign last week approximately 45,000 net counts provided SNR of about 155 and uncertainty of around 0.006 magnitudes just for the raw magnitude measurement of the target (error of the comp star NOT included in quadrature) 10,000 counts will provide an uncertainty greater than 0.010 magnitudes excluding error contribution from the comp. The importance of signal to noise is emphasized in the alert but the count recommendation provides poor SNR. It does not make sense to me to have individual image uncertainty of over 0.01 when you are trying to measure a signal of 0.002 ( or less) and you only have around 10 cycles of signal to measure in a single observing window. Is there any chance the what is intended is 10,000 net counts [average per pixel - sorry didn't write this correctly what I intended is] an average of 10,000 max pixel count for the target without saturating the comp star rather than 10,000 counts per frame?
Further, the requirements as stated seem conflicting:
"Requirements are:
i) at least one of Johnson B and V filters (both if available, cycled BVBVB..., and/or BVRBVR... or BVIBVI...)
ii) 10,000 target counts per frame (for LS IV-14 116, unless bright comparison is saturated and no other good comparisons in field). The 10,000 counts requirement is indicative rather than a minimum, and exposures of up to 100 s are acceptable, so long as the overall cadence provides a data point in each filter every 300 s (5 minutes) or less. The shortest period exhibited by the star is 1950 s, which we would like to sample well.
iii) exposures times 60 seconds or less (per filter) . . . "
It would seem to me that on the one hand exposures up to 300s are allowed but then it states exposure times 60 s or less per filter are required.
Getting SNR to detect one and two millimag signals is really tough with sufficient cadence using a 0.6 meter telescope if you have really good skies and impossible if you don't. One of the members of our astronomy club captured an exoplanet transit that had a 5 millimag transit depth, and it was just barely perceptible over the noise. 2.1 millimags and 1.9 millimags seem to be quite a stretch with a cadence that probably needs to be no longer than, say, 3 minutes - around 10 samples per cycle for the shortest period. You need a substantially higher sample rate than the Nyquist frequency because you are only sampling a few cycles per observation window and you have significant noise.
I would really like some clarification because one of our club members is preparing to bang away with 40 second integrations in B per a recommendation from the PI . Sure we can stack 4 of them if it turns out we have to, but most AAVSO members will be submitting measurements based on the requirements in the alert notice and the most strongly worded requirement is EXPOSURE TIMES LESS THAN 60 SECONDS PER FILTER and that is going to give measurements with high uncertainty compared to one or two millimag variability.
The paper is attached.
Brad Walter, WBY
Dear Brad et al.
Indeed the amplitudes are small, but not quite as bad as you imply.
First, the cited values 'a' are semi-amplitudes (a is the amplitude of a sine wave).
Second, the peak to peak amplitude can reach 20 mmag or more simply because different modes can combine constructively (conversely, variability can seem very low when they combine destructively).
Third, my published semi-amplitude were in white light. They will be larger in 'B', though harder to measure - I accept.
Fourth, Fourier analysis is a very powerful tool which looks for multiple periodic signals in noisy data
... even when invisible in the original light curve ... providing the light curve is long enough.
Thus a SNR of 100 (10,000 counts) per frame should be good. I appreciate that a seeing of 2.5 arcsec may cause problems, but the field is sparse and crowding should not be too bad. In the paper you attached,
I did indeed achieve a very much higher count rate by using white light, but with 10 and 15 second integrations.
One of our current objectives is to try to measure the colour-amplitude variation .. in otherwords the difference
between the amplitudes in B and V .. which gives an indication of the actual mode of the stellar oscillation.
However, a second objective is to place the spectroscopic measurements being made with HST and VLT into the context of the light variations, so that we can interpret the phases correctly. Amplitudes measured in just the B filter can be combined from different sites, and have a physical meaning.
White light observations are also useful, but only give frequencies and phases, and not amplitudes, since "white light" varies from site to site. If you would prefer to attempt the observations in white light, I will be able to use them, and it may give more satisfaction to your observers to see the light variations for yourselves.
You are concerned about the short exposure times. These are determined by the need to complete a cycle of filters in a period much shorter than the pulsation period (1950s). If we observe several filters, then this clearly needs to be shorter than about 1/10 of the period or 195s. If you prefer to obseve in only one filter, or in white light, then a longer exposure could be permitted, but I would suggest not exceeding 180s.
I really appreciate the enthusiasm of you and your members to contribute to this project. I have not worked with AAVSO before, so this is new to me as well, and so I apologise if I am not wholly aware of your capabilities.
I wish you clear skies.
Simon Jeffery
Simon Jeffery tried to post the following reply had some trouble. He asked me to post it. It is a very informative reply and clears up a number of questions I had. It seems the key is having a large number of points covering a bunch of cycles to get a good Fourier analysis.
Brad Walter, WBY
****************************
WBY,
Message:
[I posted this to the AAVSO site, but it needs to be moderated, so I am sending directly -- May I say that I really appreciate your feedback and questions and your efforts to participate in this project.]
Dear Brad et al.
Indeed the amplitudes are small, but not quite as bad as you imply.
First, the cited values 'a' are semi-amplitudes (a is the amplitude of a sine wave).
Second, the peak to peak amplitude can reach 20 mmag or more simply because different modes can combine constructively (conversely, variability can seem very low when they combine destructively).
Third, my published semi-amplitude were in white light. They will be larger in 'B', though harder to measure - I accept.
Fourth, Fourier analysis is a very powerful tool which looks for multiple periodic signals in noisy data ... even when invisible in the original light curve ... providing the light curve is long enough.
Thus a SNR of 100 (10,000 counts) per frame should be good. I appreciate that a seeing of 2.5 arcsec may cause problems, but the field is sparse and crowding should not be too bad. In the paper you attached, I did indeed achieve a very much higher count rate by using white light, but with 10 and 15 second integrations.
One of our current objectives is to try to measure the colour-amplitude variation .. in otherwords the difference between the amplitudes in B and V .. which gives an indication of the actual mode of the stellar oscillation.
However, a second objective is to place the spectroscopic measurements being made with HST and VLT into the context of the light variations, so that we can interpret the phases correctly. Amplitudes measured in just the B filter can be combined from different sites, and have a physical meaning.
White light observations are also useful, but only give frequencies and phases, and not amplitudes, since "white light" varies from site to site. If you would prefer to attempt the observations in white light, I will be able to use them, and it may give more satisfaction to your observers to see the light variations for yourselves.
You are concerned about the short exposure times. These are determined by the need to complete a cycle of filters in a period much shorter than the pulsation period (1950s). If we observe several filters, then this clearly needs to be shorter than about 1/10 of the period or 195s. If you prefer to obseve in only one filter, or in white light, then a longer exposure could be permitted, but I would suggest not exceeding 180s.
I really appreciate the enthusiasm of you and your members to contribute to this project. I have not worked with AAVSO before, so this is new to me as well, and so I apologise if I am not wholly aware of your capabilities.
I wish you clear skies.
Simon Jeffery
************************
I would like to thank all observers who have participated in this campaign. We had excellent skies at the VLT and the radial velocity data look as though they will be superb. The value of these will be greatly enhanced by your photometry. Although your data may not show an exciting large-amplitude pulsator ... I am hopeful that the combined volume data from several sites will allow us to extract the principal periods, phases and amplitudes.
I have one last favour to ask. We would like to analyse all of the CCD photometry, from whatever source, using a uniform software treatment. In order to do this, we would ask observers to provide us with your raw CCD frames and and whatever calibration files (dark, flat, bias) you have available.
There are two straightforward options:
1) Upload directly to a server at the University of Delaware using sftp or scp. Instructions are attached. Email me (csj -at- arm.ac.uk) in case of any difficulty.
- OR -
2) Place your files in a space where we can access them, and give us instructions on what and where to collect.
Sincerely, Simon Jeffery