EXOSAT

• lcurve: is a task which takes an input file and plots an appropriately binned light curve. The simplest way to use lcurve is to type lcurve at the Linux prompt. After loading, FTOOLS will step you through the inputs to lcurve. The inputs are as follows: number of input files, name of file, name of outside window file (which we will always denote `-'), newbin time (i.e. time bin width), number of newbins, output file name, and plotting suggestions.

  After lcurve loads the file, you will see some information about the file, including most importantly, the start/stop time of the experiment, and the bin width, usually in seconds. Clearly, the newbin width can't be smaller than the data resolution, i.e., smaller than the bin width of the input file. How do you select the number of bins to plot? To get a full screen plot, make sure that newbins $\times$ number of new bins = duration of experiment in seconds. For instance, the file d05853.lc has data collected for nearly 4 hours, or about 13920 seconds. If the bin size newbin is 30.5 seconds, make the number of bins = 13920/30.5, which is about 470. Varying newbins and num of newbins gives you a chance to best view the data.

  $\circ$ Exercise 2: Look at the light curve and power spectrum of some data taken of the source AM Her. The file is called d05479.lc (sequence number 85, broadband EXOSAT ME). This is a nice, dramatic source once you get around the fact that most of the data taken over nearly a day has zero counts.

  Enter the following parameters when prompted:
  Number of time series for this task: 1
  Ser. 1 filename : d05479.lc
  name of window file: -
  Newbin Time: 30.
  Number of newbins per interval: 20000
  Name of output file: default (will automatically make d05479.flc)
  Plot your results?: yes
  PGPLOT device: /xw.

  First try number of newbins = 20000. You'll get a squeezed plot. Control-c out of lcurve, and try again with newbins = 6000, then 3000. At 3000, you should see the data now. The interesting part of the curve is between $0$ and $1\times 10^{4} seconds$. To see error bars on the light curve, the x axis should be smaller than the entire data set. Use, for instance, number of newbins = 2000. Estimate the dominant frequency.

• powspec: These light curves have some (perhaps not very simple) periodicity. It's important for astronomers to be able to determine any periodic behavior because this may offer insights about the underlying physics of the X-ray source. The set of periodicities present in the spectum plotted with respect to intensity (i.e., strength of signal vs. frequency) is called a power spectrum. powspec returns the power spectrum for a given light curve. powspec asks for essentially the same information as lcurve.

  Figure: Gk Per power spectrum (top), and concatenated power spectrum (bottom). Note that in the concatenated power spectrum, the peak near zero is smaller relative to the one near $3\times 10^{-3}$.

  

• Concatenation. This is not a command but an extremely useful technique to use with powspec. You can use this two ways:
  1. You can add the data sets together to increase the amount of signal. This may or may not also increase the noise levels. If the noise levels are due to some systematic error, then the noise level will also increase. It the noise is due to random error, then the noise may have a much lower level than the signal. If the noise levels are tolerable, then you will get a much better idea of the data set's characteristics.
  2. There are situations where the lightcurve you're studying has so much noise that the peaks in the powerspectrum that are genuine signal are hard to distinguish from those due to random noise because they are about the same height. One way to fix this problem is to combine the data sets while averaging different times. The resulting power spectrum is the average of the power spectra from each data set. A real signal will likely appear in all sets, and so the average for that signal will be strong. Random errors will appear in some, but not all data sets, and will have different frequencies in each, so the average at at any one frequency will be very weak. Therefore, the signal will look significantly stronger relative to the noise after concatenation than it would before.

  So, how do we concatenate? Just do the following:

  1. Run lcurve on each file (i.e. each data set) to generate a .flc file for each

  2. Place all .flc files, one per line, in a text file, for example, `powcat.txt'.

  3. Start powspec and type @filename (@powcat.txt, in our example), when prompted for the file

  4. Define the number of intervals per frame:
     • Choose 1 (usually the default), if you'd like a true concatenation (adding) of the data. This is used to increase the amount of data (which may include both signal and noise) so that you can more easily visualize the data.
     • Choose N where N is the number of files in the data set. This concatenates by averaging.

  $\circ$ Exercise 3:

  Use powspec on the same file as exercise 2, (d05479.lc) and compare your estimated frequency to the calculated frequency.

  Use the following parameters when prompted:
  Ser. 1 filename : d05479.lc
  default window: -
  Newbin time: 30.
  Newbins per interval: 2000
  Intervals per Frame: 1
  Rebin results?: 0
  Output file: default
  Plot results?: yes
  PGPLOT device: /xw.

  Note the huge peak to the far left of the plot. Note also the PLT$>$ prompt and realize that you can use something like rescale x 0. 1.e-3 to rescale the plot. Determining the peak frequency will be impossible without rescaling.

  $\circ$ Exercise 4: Look at the power spectrum of the another measurement of AM Her. These files (EXOSAT ME sequence no. 167) are located in the directory
  tutorial/ftools/amher/ and are named d08040.lc and d08054.lc. First look at the lightcurves using lcurve, and then look at the spectrum of a single source (just choose one), and then examine the spectrum when you look at both light curves. Use concatenation to find the power spectrum of both light curves.