Re: GSC 00279 00321 and a quicky AVE tutorial

[jg <jg@jgws.freeisp.co.uk>]

Michael Koppelman wrote:
> 
> OK, I got some more data on this. It is here:
> 
> http://www.lolife.com/gsc279-321/
> 
> 20020531.jpg  Light curve from first night
> 20020601.jpg  Light curve from second night
> combined.jpg  Light curve of both nights
> combined.txt  Data from both nights
> 
> The period seems to be around 0.128 days. I'm going to try to get a phase
> plot and an ephemeris. I'm also going to cross-check other sources to get an
> idea of the color. Between my data, the TASS data and other sources we
> should be able to get close to figuring this one out.
> 
The strong very nearly one day period alias is giving me headache, but I
think it is likely to be a period that is going to be near a "small
integer fraction" of one day, ie 1/2 , 1/3, 1/4, 1/5 etc, except a bit
off from that coz it isn't quite a day between the tips of your two
troughs.

0.128 is near an eighth, but I can also get quite a nice plot with 0.17,
near a sixth.  Note that the light curve probably isn't completely
sampled yet, you can't see one complete cycle even if you concatenate
the different nights, so be suspicious of a folded lightcurve that is
totally filled in.

Don't be afraid to combine it with the V data from CD 23, the offset
only looks to be about 0.25 mags, and try again.  If this doesn't fold
nicely, you've got problems.

Okay Mike, and whomever, Quicky AVE tutorial:-

Make the data two columns separated by a single space.  Remember to save
the data with sufficient decimal precision for all entries if exporting
from a spreadsheet.  Crop the preceeding "24" on Julian data before
analysis.  Probably programs designed to do this sort of analysis have
enough precision anyway, but 24nnnnn.nnnnn is a lot of significant
figures, and some programs can have difficulty just adding up when it
comes to such things, and it's the important (in this case) decimal end
that gets lopped off!

Load file into AVE.

Go to the menu bar and select TOOLS -> Period Searching

Right click on the graph, select Search Period.

Select a period range, roughly guesstimateable by looking at a plot of
the raw light curve.. ..if you've got a trough in it about half a day
wide, the period isn't going to be .1 day.  Ideally you're supposed to
use the "Nyquist frequency" for minimum possible period and half the
data duration for longest possible period.  I forget the exact Nyquist
frequency definition, but in terms of period it is twice the period of
the minimum interval between observations.  Often the mean interval
between observations is acceptable for unevenly spaced observations. 
(Actually, it's possible zero is supposed to be the minimum, I don't
remember).  Points, well if computer is fast, a 1000 points should do
it.  Ideally I think it should be half duration over 1/(Nyquist
frequency)... ...not sure on that bit at all, though, as normal practice
is to set a frequency step size, which would be the Nyquist frequency
again (ie, how can you find a period that exists between the smallest
inter-measure interval).  During analysis you'll get used to people
quoting frequency one moment and period the next whilst looking for
periods, it's just the way it is, just remember frequency = 1/time.

In your case I'd generally look between 0.1 and 1.5 days to start with,
just as a test.

For analysis choose either PDM, which works okay with non-sinusoidal
lightcurves, though is tricky to interpret, or use, as in this case, the
Periodogram(Bloomfield) option, mostly because the output is nicer in
this package with that one!  The periodogram options are all different
versions of "DFT" using different methods of weighting to avoid aliasing
problems etc and I think sometimes trying to weight via statistical
tests.  (DFT is Discrete Fourier Transform... ...try not to use packages
that use Fast Fourier Transform on lightcurves, as this only works
properly on continuous, evenly spaced data, apparently).

I choose a few likely peaks (drawn as troughs for DFT plots in AVE for
some reason) and re-analyse on a shorter range of periods.

Then for each of these you right click _when the cursor is placed
exactly over the place at the tip of a trough you feel represents exact
midpoint_ thus giving the assumed period (you can read it off the status
line).  Anyway, you right click and select Mark Period for this point. 
That period is marked now so when you right click again you can choose
phase diagram and a phase diagram is plotted at that marked period.

You stare at that, and it stares back.  You carry on, and if enough of
the light curve is sampled you should get a pretty good light curve at
some point from one of the periods... ...actually that period should
stand out better than others.

Finally, for some publication work it can be a lot more involved than
that, using sledgehammers like searching the entire possible frequency
range, something else called spectral window to check for aliasing etc,
&c, even if the nut has already been cracked.  All that is just to show
there's not another solution, or that there are no other periods.  Don't
know how often this is really done in reality though.  From earlier
postings I assume stuff like IRAF and what have you pretty much
automates a lot of this.  Only spectral window capacity I've got is an
abysmal bit of home grown qbasic code that I've had to leave running
overnight for large datasets before now.

Any muddly bits ask (there'll 18h time lag, I'm going to bed!), anybody
wanting to clarify any of the above where I've been a bit loose with the
description or jargon, feel free to.  I am out of practise and usually
deal with longer period variables than these, with relatively longer
data runs.


Cheers

John