The time span of some data sets is too large to perform a single fft to search for short periods. Some ROSAT observations contain gaps of days length, for example. The data can be divided into a set of individual intervals, all of the same length. The fft will transform each interval and sum the power coeff for all intervals. Gaps can be eliminated or minimized by clever selection of interval length. However, the shorter the interval the less sensitive the transform of the data set.
First make a file (e.g., rp90_sum.lis) which defines the .qp files of each data interval to be used as input to the transform,
| e.g. | rp90_sti.qp[time=(1606720.0:1607232.0)] |
| rp90_sti.qp[time=(1611613.0:1612125.0)] |
Note all intervals are 512 sec long in this example. FFT is the only timing task that will take more than one .qp file as input.
Now run the FFT
st> fft
input source file (table or time_sorted qpoe file): @ rp90_sum.lis
# the ``@'' is important
rootname for output file [root_fft(ftp,pwr).tab]: rp90_sum
input bgnd qpoe file (name): <cr>
input table column heading (src): <cr>
number of bins: 0 # a power of 2
bin length: .025 # important to specify bin length when intervals
# are not full of data (contain gaps)
The screen will report each transform as it is completed and the
generation of the _fft, _pwr, _ftp output tables.
Note that as long as all intervals do not contain gaps (regions of
bad data) at start and/or end, fft may be run with number of bins
specified. If there are gaps at start and/or end of any interval, the
bin length must be specified.