There are five stages to the search process
The first step, extraction, involves getting the raw data and modifying the data and headers into a form that can be read by sigproc. This step requires different software depending on the telescope and hardware used to record the data.
The second step, dedispersion, uses sigproc’s dedisperse program to reduce the multi-channel data to a single-channel data file, corrected for a given dispersion.
The Fourier search step is where the FFT and associated processes occur. This is where the candidates are first detectcted.
Next, the detected signals are collected together to make individual candidates, which then can be optimised in the time domain to tune the parameters and maximise the signal-to-noise ratio.
Basic usage:
$ dedisperse [infile].fil -d [dm] > [outfile].tim
This produces a single channel time-series file de-dispersed at the given DM.
Additional options:
-b [n] -nobaseline
produce a n-channel time series output. Each of the n channels is de-dispersed to the top of that band, however the chanels are not dedispersed with respect to each other (i.e. you need to de-disperse again to get a single channel time series, rather than just Fscrunch).
Note, that for all known sigproc versions up to time of writing (up to 4.3), and probably many future versions, creating sub-bands seems to cause bad data if not used with the -nobaseline switch!
Fourier searching in sigproc is done with the seek program.
Seek produces a .prd file, listing the periods and signal-to-noise ratios that have been detected. This can be read with best, or ph-best.
Example useage:
$ seek [infile].tim -pulse -recon -head
Description of the options:
Other useful options:
-a[acn] -d[adot]
remove acceleration and accelearation deriviteives of acn m/s/s and adot m/s/s/s. Note that there is no space between the -a/d and the acceleration/adot value.