Next: How to Use SPECTRAL Up: User's Guide to SPECTRAL Previous: User's Guide to SPECTRAL

Program Method

The essential idea of SPECTRAL is the following: take a set of energy bins extending across the EGRET range, and place the observed counts in these bins. If necessary,can make an estimate of how many counts in each of these bins were due to off-pulse background emission instead of the point source; subtract off the background. For any hypothetical energy spectrum of s incident upon the detector, we can calculate the corresponding average observed spectrum. We pick a model for the incident spectrum, such as power-law, broken power-law, or continuum with lines. This model will have several free parameters. We can adjust these parameters until the best fit of the model observed spectrum to the actual data is achieved.

When doing pulse-phase spectroscopy, the main input to the program is a data file produced by the program PULSAR. This program produces a file in SELECT format in which events are selected from a database by their arrival times, arrival directions, energies, and other criteria. Each event has a pulsar phase value attached.

When analyzing a DC source, the data input is a file produced by PSRSPEC, which does a likelihood analysis of sky maps. This file includes the estimated (background-subtracted) flux for the source in a number of energy bands.

Other inputs include the instrument calibration files for the effective area, the energy resolution, and the angular resolution. These are used in relating the incident flux to the observed spectrum.

At present SPECTRAL is based on an iterative weighted least-squares fit to binned data. We plan eventually to replace this with a maximum likelihood method based on Poisson distributions; the essential structure and interfaces of the program will not be changed.

next up previous
Next: How to Use SPECTRAL Up: User's Guide to SPECTRAL Previous: User's Guide to SPECTRAL