rsp2rmf -- converts old-style (SF) format response matrices to OGIP FITS format
rsp2rmf rspfil rmffil
rsp2rmf reads i/p data from an old-stle (SF) format response matrix (used by XSPEC versions <8.2), and writes an o/p FITS data file in OGIP standard format (able to be read by XSPEC versions >8.2). Users are urged to switch to using XSPEC with FITS files as soon as possible.
A detailed description of the o/p FITS file format is given in
George etal (1992 Legacy, 2, 51), and in the OGIP Calibration Memos CAL/GEN/92-002 & CAL/GEN/92-002a available via the OGIP anonymous ftp account on legacy.gsfc.nasa.gov (in Postscript) as caldb/docs/memos/cal_gen_92_002.ps & cal_gen_92_002a.ps respectively. In brief, the data is written in the form of 2 BINTABLE extensions, containing:
The data contained in Extension (1) above is in a compressed format whereby only elements for which the value of the redistribution matrix above a given threshold are stored. This provides a significant reduction in disk-storage requirements for sparsely populated matrices. The default threshold is read from the i/p file, but can be altered via the hidden parameters gregrp and threshold parameters. Users are reminded that the threshold is defined in absolute units (NOT relative to the value of the maximum value in the matrix).
A further reduction in disk-storage space requirements can often be achieved by the use a variable-length array for the BINTABLE column containing the matrix. In such a case, the number of elements within this column varies between different rows of the BINTABLE. The current version of rsp2rmf will automatically write a variable-length array if a saving of greater than a factor 3 in storage space is indeed achieved.
Currently only the following OGIP standards are supported by this task
(via the parameter rmfversn):
For further details see OGIP Calibration Memo CAL/GEN/92-002a.
The current version of this task assumes that the matrix elements of the input SF file have already been multipled by the total effective area of the focussing/collimating optics, filters & detector efficiency (since this is the case for all RSP files currently available within the OGIP). Thus the redistribution matrix extension has the keyword HDUCLAS3= 'FULL'. This value of HDUCLAS3 is however incorrect in cases where the redistribution matrix alone is stored, and hence the values of the HDUCLAS3 keyword of the output file should be changed appropriately (see calibration memo OGIP/92-002a)
Users are reminded that the EBOUNDS extension contains only the nominal energies associated with each PHA/PI channel. Extreme care should be exercised if these are to be used by downstream software (see George etal 1992 Legacy, 2, 51, section 3.2). The OGIP-supplied spectral fitting package, XSPEC, only uses the data within this extension for plotting purposes.
The facility to override the deconvolution of the mission and instrument strings derived from the corresponding string supplied in the SF file is often useful since the strings required to specify for which mission, instrument, detector and/or filter the dataset is valid in the case of the output FITS file are often very different to those used in the SF file. Specifically, the SF RSP files often contain non-OGIP-standard strings, and/or insufficient information. A list of OGIP-standard strings can be found in the OGIP Memos OGIP/93-013 available via the OGIP anonymous ftp account on legacy.gsfc.nasa.gov (in Postscript) as caldb/docs/memos/ogip_93_013.ps. The current version of the task only performs a rather crude deconvolution.
In order to handle the biggest matrices, this task employs a number of relatively large internal arrays. These arrays are dynamically allocated, but problems can arise at execution if the local machine runs out of swap space. When this occurs, often an appropriate system error message will be displayed or the program may simply crash. Under these circumstances users will have to free up space by closing unwanted applications (windows, tasks etc). Under unix/ultrix, the total used/available swap space on the local machine can be listed (in kbytes) using the command 'pstat -T'.
The largest matrix able to be handled by the current version of this task is 4096 channels by 2048 energies (requiring 35 Mbytes of swap space). Due to dynamic memory allocation, the swap space requirements of most matrices are dramatically smaller.
None known
CAL/GEN/92-002 (George etal 1992 Legacy, 2, 51),
CAL/GEN/92-002a
Ian M George
HEASARC
NASA/GFSC
http://heasarc.gsfc.nasa.gov/cgi-bin/ftoolshelp
(301) 286-6094