A ``long'' FFT

Compute FFT from the input QPOE file.

input file: rp90_sti.qp

output files: rp90_ftp.tab, rp90_pwr.tab

In searching for short periods using a long data interval many bins are necessary [(no. of bins) = $2\times(data\:interval)(min\: period)^{-1}$] and generation of a _ltc.tab file may prove impractical. In such cases, the QPOE file may be input. Only the largest Fourier coefficients, but not the complete list of (e.g., $5 \times 10^{5}$) coefficients should be used, so we set power_thresh to exclude all but the largest from the table [root]_pwr.tab. The normalization has been adjusted so the value of the power coefficient, normpower, does not depend strongly on number of bins or number of counts. A significant signal will have a value of $\sim 4$ for a 10³ bin transform and a value of $\sim 8$ for 10⁵ bins. Time required for a 10⁵ bin transform is 2-3 minutes.

  xt> fft.fftcoeff=no              # don't generate a long list of coefficients
  xt> fft.power_thresh=7                       # set the significance threshold

  xt> fft

  input source file (table or time-sorted qpoe file):  rp90_sti.qp
  root name for output files [rrot_fft(ftp,pwr).tab]: rp90lng
  background qpoe file: <cr>
  input table column heading:  src
  number of bins 2048 # The task uses the closest power of 2 to the input value

FFT done with 2048 bins!

  creating histogram output file rp90_ftp.tab
  creating fft power output file rp90_pwr.tab