Data Format: Type II - multiple datasets

For a number of analysis tasks in which multiple PHA datasets are to be stored and/or analysed simultaneously, it is generally considered cumbersome if each resides in a separate file. This is particularly the case for (say) sub-second, time-resolved spectroscopy when a single observation may give rise to a very large number of datasets. It should be stressed however that this is for datasets from the same instrument. Users wishing to simultaneously analyse PHA data from different instruments should always use multiple input files. Finally, it is worth noting that users who prefer to use the multiple file approach for such analysis (especially if only a small number of datasets are to be considered) can still do so by using the XFLTXXXX keyword described above.

As in the single dataset case, the data is stored as a BINTABLE extension. However here, vector arrays are used in place of scalars for several columns. Thus the PHA data, errors etc. are vectors and each row contains a single PHA dataset. The other columns list other parameters which may vary between the datasets. The notation below is the same as that used in description of the type I data format.

1. Num, a 2-byte INTEGER scalar giving the reference number of the spectrum stored in this row.
   The FITS column name is SPEC_NUM.
   (unitless).
2. Channel, a 2-byte or 4-byte INTEGER array containing the channel numbers for the spectra in the Data column.
3. Data, either:
   • a 2-byte or 4-byte INTEGER array containing the number of counts observed in all channels during this time interval;
   • or a 4-byte REAL array containing the number of counts per second observed in all channels during this time interval.
   In both cases the notation is as for the type I data.
4. Stat_err, a 4-byte REAL array giving the statistical error on the values within Data.
5. Sys_err, a 4-byte REAL array giving the fractional systematic error to be applied to each value within Data.
6. Qual, a 2-byte INTEGER array giving the data quality flag for each value within Data.
7. Grpg, a 2-byte INTEGER array giving the data data grouping flag for each channel.
8. Rowid, an (optional) CHARACTER string giving a unique identifier for each spectrum. The FITS column name is ROWID.
9. Exposure, a 4-byte REAL scalar or array giving the exposure time for each spectrum.
   The FITS column name is EXPOSURE
10. AreaSc, a 4-byte REAL scalar or array giving the area scaling factor to be applied.
    The FITS column name is AREASCAL
11. Bkgdfile, a CHARACTER string giving the name of the background file (if any) associated with the dataset contained in Data.
    The FITS column name is BACKFILE
12. BackSc, a 4-byte REAL scalar or array giving the scaling factor to be applied to Bkgdfile.
    The FITS column name is BACKSCAL
13. Corrfile, a CHARACTER string giving the name of the correction file (if any) associated with the dataset contained in Data.
    The FITS column name is CORRFILE
14. Corrscal, a 4-byte REAL giving the scaling factor to be applied to Corrfile.
    The FITS column name is CORRSCAL
15. Respfile, a CHARACTER string giving the name of the redistribution matrix file (RMF) associated with the dataset contained in Data.
    The FITS column name is RESPFILE
16. Ancrfile, a CHARACTER string giving the name of the ancillary response file (ARF; if any) associated with the dataset contained in Data.
    The FITS column name is ANCRFILE

Points to Note & Conventions (in addition to those given in the section on type I data)

• In addition to those listed the following keywords may appear as columns : XFLTXXXX, DATE-OBS, TIME-OBS, DATE-END, TIME-END.

• As with all OGIP files, any of the above variables which have the same value in every row can be removed from the table, and specified as a keyword within the header. Here this is most likely in the case of EXPOSURE, AREASCAL, BACKFILE, BACKSCAL, CORRFILE, CORRSCAL, RESPFILE, and ANCRFILE.