Gtpsearch Tutorial

Introduction

The gtpsearch tool searches for pulsations in data which is known or suspected to have a pulsation of a known approximate period or frequency. It is not useful for a so-called blind period search, in which data are examined for pulsations at any frequency.

Prerequisites

The only required input file is an events file. The user also must have information about the suspected pulsation. This information can be input directly by the users for many cases. If the pulsation in question is believed to be associated with a known pulsar, an optional pulsar ephemerides database file can also be used to supply the necessary search parameters. A standard pulsar ephemerides file is available alongside the gtpsearch tool in the periodSearch/data directory. Files of this type can also be created/edited/merged using the gtpulsardb tool.

How To Perform A Simple Search

The following sample analysis uses the file step-01.fits, which is included in the periodSearch package. This file is a standard GLAST events file whose TIME column contains ASCA data for the observation of PSR B0540-69, or the 50-ms pulsar in the Large Magellanic Cloud. For purposes of this example, suppose the best estimate for the frequency is 19.8339 Hz. The epoch is taken to be 23078385.922, which is the center of the observation. (The meaning of epoch and its effects on the computation as well as the other parameters will be explained in detail below.)

First Try, Using Central Period Only

The full output produced by this command is quite long, but the first few lines are:
A plot will also be produced which is similar to the following plot.

gtpsearch example 1

Second Try, Correcting For Known Frequency Variations

From the shape of the graph, it appears that there is a strong maximum with mainly symmetric side lobes. Encouraged by this result, one might run the tool again, using the position of the peak (stated at the top of the tool output) as the central frequency. To zoom in on the peak more closely, the number of trials will be reduced to 100. In addition, the process of correcting for variations in the frequency known from other observations will be demonstrated. In this case, the frequency derivatives were taken from an analysis of the same ASCA data.

gtpsearch example 2

The effect of correcting for frequency variation was to increase the value of the maximum statistic, or the height of the main peak. Changing the central frequency centered the graph on the main peak. With the smaller number of trials, the second lobe on either side was cut off.

Third Try, Looking For Precise Centroid Of The Peak

By decreasing the steps between subsequent trials, one can see the shape of the peak in a finer resolution.
gtpsearch example 3

How To Use Other Statistical Tests

In the examples above, the Chi-Squared statistic was used for purposes of finding the pulsation. The gtpsearch tool currently supports two other test statistics: the Z2n test and the H test. A full explanation of the strengths and weaknesses of these approaches is beyond the scope of this document, but in brief, the most effective test depends on the pulse shape. For the sample data used in this test, all the tests return similar results, so a specific example will not be shown.

Known Issues

Meaning Of Epoch

The epoch is a somewhat arbitrary time origin which is subtracted from the times of the events. Its exact value is important only when frequency variations are being taken into account, because the correction estimates the frequency by multiplying the derivatives by powers of the difference between the event time and the epoch. In most cases, the center of the observation time, (TSTART + TSTOP) / 2, is the best choice for the epoch.