next up previous contents
Next: 10. Spatial Analysis Up: ASCA ABC Guide Previous: 8. SPECTRAL ANALYSIS   Contents

Subsections


9. TEMPORAL ANALYSIS

9.1 Introduction

This chapter concerns extracting light curves after the data have been screened and/or filtered to prepare the data for temporal analysis. There are three formats for light curves - two FITS and one ASCII. Also covered in this chapter are dead time, barycentric and geocentric corrections and getting started with the timing analysis program XRONOS. The User's Guide can be obtained at the URL

http://heasarc.gsfc.nasa.gov/docs/xanadu/xronos/xronos.html.

For users who want to produce light curves for figures rather than for XRONOS-style temporal analysis, we show how to use XSELECT for creating light curve figures.

The latest information about the accuracy of ASCA timing analysis can be found in the `Calibration Uncertainties' Web page (§1.7).


9.2 Three Formats of Light Curve and How to Extract Them

Light curves are produced in XSELECT either for creating time filters, i.e., for extracting events based on time, or for temporal analysis outside XSELECT. For this latter purpose, the main subject of this chapter, XSELECT can create light curves in three formats, each of which has its advantages and disadvantages:

  1. Events lists (FITS),

  2. Rate files (FITS) and

  3. QDP files (ASCII).

Although their arrival times are quantized with the precision appropriate to their data mode and bit rate, the events in an events list should be regarded as unbinned. On the other hand, a light curve is ipso facto always binned. To derive the correct count rate in each light curve bin, exposure information is required so that the absence of counts can be attributed correctly either to the non-detection of a photon or to the instrument not being on. When extracting light curves from ASCA events lists, this means making sure that when the start or end of a GTI falls within a bin, the counts in that bin are reduced in the same proportion as the bin is shortened. This issue is dealt with differently depending on the light curve format.

With light curves, the most common filter used is an energy filter which allows light curves to be extracted in various energy bands.

Errors for the FITS rate and QDP files that are not background-subtracted are calculated in the same way, by taking the square root of the number of counts in the bin and dividing by the binsize.


9.3 Event Lists as Light Curves

The timing analysis program XRONOS can read in the screened GIS and SIS events lists. When the program bins up the data, it consults the GTI extension to calculate the correct exposure. With the original events list (or a region-selected subset of the events list), users always obtain the highest (i.e., the original) temporal resolution. However, the events list contains a lot of information (such as PHA and position) that XRONOS does not need, so the input file is rather large.

In XSELECT, to produce an events list from the region of interest, use the command extract events followed by save events, after applying the desired filters.

The structure and contents of events lists, i.e,.the basic FITS data files, is covered in §2.


9.4 FITS Rate Files

FITS rate files contain the light curve in the first extension (called RATE), and the good time intervals in the second extension (called GTI). They differ from events lists in two important respects: they are binned and they contain only rate data (i.e., no columns for position, grade, RTI, PHA, etc.). This results in a considerable saving. For example, an events list containing SIS 1-CCD BRIGHT mode data from GX301-2 occupies 6.6 times more disk space than the corresponding rate file binned at the minimum 4s resolution of the data.

XRONOS can read this file and calculates the exposure correctly by applying the GTIs to the binned counts.

In XSELECT, to produce a FITS rate file from the region of interest, first choose the binning with the set binsize command. Then use the extract curve and save curve commands to extract, save and name the file. NOTE: If the screened events list from which you extract the light curve contains data with different temporal resolutions (e.g., GIS PH mode with MEDIUM and HIGH bit rates), then make sure that the binsize you choose is no smaller than the largest in the data.


9.5 QDP (ASCII) Rate Files

Many users simply want a binned light curve - with errors and in ASCII - which they can read into their own timing analysis or plotting programs. The QDP rate file satisfies this need. Apart from an ASCII header, the file comprises columns containing time, count rate and count rate error. It can be read directly into the QDP plotting program, as well as into XRONOS. The exposure information that XRONOS needs is taken from an accompanying file (prefixed with B) that XSELECT produces when the QDP rate file is saved. The procedure for producing and saving the QDP rate file is the same as for the FITS rate file except that the commands are augmented with the appropriate qualifiers - like this (for a light curve called 3c273_lc_sis0.qdp with 32 s bins):

xsel> set binsize 32
xsel> extract curve use_qdp = yes
xsel> save curve save_qdp = yes 3c273_lc_sis0.qdp


9.6 Creating Light Curve Figures with XSELECT

Users not interested in full-blown temporal analysis might still want to produce a light curve for a figure. Fortunately, XSELECT has an interface to the PGPLOT graphics package which can be used to produce plots without having to quit XSELECT and run another program. The same interface is used when creating time and intensity filters.

To make a light curve figure from screened and filtered data execute the following scheme. An alternative is to edit and insert the commands directly into the .qdp or .pco file.

  1. Set the size of the light curve bins with the set binsize command. The binsize in seconds is the argument.

  2. Extract a light curve with the extract curve command.

  3. Plot the light curve with the plot curve command. If you have not already set a plot device, then you will be prompted for one: /xw, for XWindows, is a good choice for most workstations.

  4. A plot will now appear on the screen with the default appearance. To tailor the plot to your requirements, enter commands at the PLT$>$ prompt. Here are some of the basic ones. To see a full list, type ? at the PLT$>$ prompt. Note that some commands, like rescale, automatically re-plot, while others, like label, require an additional plot command to see the results of a change.

  5. The final step is creating a hardcopy which is done with the hardcopy command. For example, to create the PostScript file, eclipse.ps, type hardcopy eclipse.ps/ps at the PLT$>$ prompt.

To do more sophisticated manipulation consult the QDP/PLT manual and/or the on-line help.

9.7 Background Subtraction for Light Curves

There is no general program for subtracting background from ASCA light curves. However it can be done with a combination of FTOOLS or your own software. There are two basic methods. Both methods require making a light curve from a background region from the same observation, with the same selection criteria and preferably the same GTI.

  1. If the binsize is large enough that you have enough background in each bin, subtract the background from each corresponding on-source light curve bin directly.

  2. Since the ASCA background will be fairly constant during an observation, especially after screening, you can fit the background light curve with a simple function such as a polynomial (and in most cases, simply a straight line) and subtract this from the on-source light curve.


9.8 GIS Light Curve Deadtime Correction

GIS PH mode data has a typical deadtime of about 8 msec per event, and the deadtime has to be corrected for bright sources in order to obtain the intrinsic X-ray flux and its time variation. GIS deadtime can be calculated using the GIS monitor count rates, and the mkf files created in the standard processing (REV1 and later) have two columns, G2_DEADT and G3_DEADT, in which the deadtime fraction values thus calculated are stored.

The FTOOL ldeadtime uses these deadtime fraction and the deadtime correction can be done as follows. In this example of the use of ldeadtime the GIS light curve is called g2.curve which has a corresponding mkf file ft941123_0821_2301.mkf:

ldeadtime g2.curve ft941123_0821_2301.mkf

The RATE and ERROR columns of the light curve file are overwritten, now corrected for the deadtime. The FRACEXP column will have the livetime fraction for each time bin, and can be used to get obtain the original counting rates before the correction.


9.9 Telemetry Saturation Effects for SIS Data

Since the SIS is not a true photon-counting detector, there is no hardware deadtime as such: a pixel into which an X-ray photon enters remains responsive to further X-ray photons throughout the exposure cycle. This does, however, lead to the problem of `photon pile-up' for the spectroscopy of very bright sources.

In time-series analysis, the most important SIS `dead time' effect is telemetry saturation. When the total count rate of the source, X-ray and non X-ray background exceeds the telemetry limit of the SIS mode/bit rate combination, events are lost in a manner that makes dead-time correction difficult. For mkf files originating in REV1 or later processing, a series of telemetry saturation flags (Sn_SATFm, where n is the SIS number and m is the chip number) are included. By selecting, say, S0_SATF1 = 0, you can select time intervals when events in SIS-0 Chip 1 did not saturate the telemetry.

Telemetry saturation limits, in counts per second per chip, are summarized in the following table.

        ----------------------------------------------------------
                     FAINT mode       BRIGHT mode       FAST mode
        ----------------------------------------------------------
        CCD mode     1    2    4       1    2    4          PS
        ----------------------------------------------------------
        High        64   32   16     256  128   64         512
        Medium       8    4    2      32   16    8          64
        ----------------------------------------------------------

The rate of hot/flickering pixels has to be included when assessing the net count-rates as well as the source and the X-ray background count-rates.


9.10 Barycentric and Geocentric Corrections

The FTOOL timeconv converts photon arrival times (represented by satellite time by default) to geocentric or barycentric time. It was written by Yutaro Sekimoto and Masaharu Hirayama (University of Tokyo), and converted to an FTOOL at GSFC.

The program timeconv runs on the screened events list rather than on the FITS rate files or QDP rate files. Note that since it alters the times in the mkf file, timeconv should be run after all selections have been performed. Two files are needed.

The program timeconv is run as follows:

$>$ timeconv data.evt m ra dec earth.dat frf.orbit

where data.evt is the screened events list; m is set to

RA and DEC are self-explanatory (and must be given in J2000).

Since the time systems used in geocentric time and barycentric time are different, care must be taken when using timeconv. The following is a brief description of several different time systems:

The three options in timeconv work as follows:

  1. Geocentric: Correct the light travel time between the satellite and the center of the Earth.

  2. Barycentric: Correct the light travel time between the satellite and the barycenter of the solar system, and then convert the photon arrival time, which is represented by ASCATIME, to TDB. Note that TDB differs from ASCATIME by 32.184+27.0 sec. Timing results from different missions are conventionally compared by using TDB.

  3. Barycentric-ascatime:Useful to see the time difference between the satellite and the barycenter.


9.11 GIS Light Curves Using Monitor Counts

There are GIS house-keeping parameters which are called `monitor counts' and can be used to study time variations of X-ray sources. Among several different kinds of monitor counts, the `L1' monitor counts are considered mostly X-rays, and suitable for astrophysical study. Although the L1 monitor counts do not have energy nor spatial resolution, its small deadtime (25 micro-sec/event), large telemetry limit (2048 cts/s/sensor) and moderate time resolution (0.125 sec) make it effective for timing studies of bright X-ray sources for which standard PH mode data might be hardly usable because of the significant deadtime and the telemetry saturation.

Here is an example of making a FITS format L1 light curve, g3_L1.curve, from the HK file, ft950420_1006_2240G3HK.fits, using the FTOOL ghkcurve:

ghkcurve ft950420_1006_2240G3HK.fits g3_L1.curve bit_high.gti High

bit_high.gti is the GTI file. The last option is either HIGH or MEDIUM, corresponding to whether the GTI file includes HIGH bit rate (with a time resolution of 0.125 sec) or MEDIUM bit rate (with a time resolution of 1.0 sec) respectively. The light curve file thus made can be read by XRONOS in the standard manner.


9.12 Temporal Resolution

The temporal resolution of ASCA data depends on instrument, mode, and - in the case of the GIS only - bit rate.

The keyword TIMEDEL contains the temporal resolution of an events list, i.e., the smallest data accumulation time. It appears in the primary header and in the header of the EVENTS extension (but SPECTRUM extension for MPC mode). But beware: when events lists of differing temporal resolution are combined (e.g., when combining GIS HIGH and MEDIUM bit rate data), TIMEDEL has the value of the first set of events, regardless of whether it applies throughout the file.


9.12.1 GIS PH Mode

For standard bit assignments, the temporal resolution is 62.5 msec in HIGH bit rate and 500 msec in MEDIUM bit rate.

For non-standard bit assignments, when N bits are used for timing, the temporal resolution is 2**N times better.


9.12.2 GIS MPC Mode

The temporal resolution in MPC mode depends on bit rate, as well as on how many bits are assigned to timing versus PHA, as shown in the table below.

        ---------------------------------------------------
                                Timing bits 
 
        BIT RATE         0        2        4        8
        ---------------------------------------------------
        HIGH           0.5s   0.125s   31.25ms   1.95ms
        MEDIUM           4s      1.s    0.25s   15.60ms
        ---------------------------------------------------


9.12.3 SIS FAINT, BRIGHT and BRIGHT2 Modes

For the three SIS imaging modes, FAINT, BRIGHT and BRIGHT2, the temporal resolution depends on the clocking mode, i.e., how many CCD chips are exposed. It does not depend on bit rate:

        -------------------------------------------------
        CLOCKING MODE	 4-CCD     2-CCD     1-CCD 
        -------------------------------------------------
                         16s        8s        4s
        -------------------------------------------------


9.12.4 SIS FAST Mode

In FAST mode events lists, the values in the TIME column are the photon arrival times with a precision of 4 s. The RAWY column contains a correction to this value which, when applied by XSELECT (via the FTOOL fasttime), increases the precision to 15.625 ms.


next up previous contents
Next: 10. Spatial Analysis Up: ASCA ABC Guide Previous: 8. SPECTRAL ANALYSIS   Contents
Michael Arida 2002-10-22