PCARMF





PCARMF is the tool which creates 256 channel response matrices for 5 individual Proportional Counter Units (PCU) which jointly comprise the Proportional Counter Array. PCARMF is called several times by PCARSP (a perl script) to create a matrix suitable for arbitrary PCA data. Depending on the mode in which PCA data was collected the channels may have been shifted to an approximately constant energy scale, the channels may have been rebinned, and data from different detectors may have been co-added. To create a suitable response matrix for PCA data, the observer must first create a matrix for each detector (using PCARMF), shift the channels if necessary (GCORRMF), create the ancilliary response file (XPCAARF), combine the redistribution matrix and the arf file (MARFRMF), rebin the results to match the telemetered channels (RBNRMF), and add the matrices together (ADDRMF). This page is designed to provide up-to-date information only on PCARMF.


August 1998
The best available matrices have been delivered with the FTOOLS 4.1 delivery, and include a modest time dependence (i.e. you should create matrices customized to your observation using PCARMF or PCARSP - see the GOF documentation or ftools help). Don't forget to read the release notes associated with ftools 4.1 and to get the updates to the CALDB. See the also the ERRATUM: posted on the PCA page. This error will be corrected in the ftools v4.2, expected to be released in late 1998.

Check also on the PCA page for status updates on matrix studies which were reinvigorated in October 1998.

1/20/98 Some of the links have been removed. Contact Keith Jahoda if you need access to the older links.

PCARMF v2.2.1
January 20, 1998. An updated set of v2.2.1 matrices with "resolution parameter=0.80" was available here. These matrices differ from the previously available v2.2.1 matrices only in the instrumental resolution, which was previously understated (i.e. equivalent widths were correct but physical widths were too large). These matrices correspond to the M80 matrices referred to in Weaver et al. (astro-ph/9712035). For comparison, the previous version of v2.2.1 matrices are still available. All of the documntation in this section still applies.
The matrices are all named pN_xx_yyyy_eM.rsp.gz

where N=0,1,2,3,4 to describe each PCU
      xx = 3 for first layer (L1+R1)
          12 for second layer (L2+R2)
          48 for second layer (L3+R3)
          63 for sum of all xenon layers
      yyyy = std2 for 129 channel binning approriate for Standard 2 data
           = 256 for full resolution (start here if you are constructing a
               matrix for data that has had EDS gain and offset corrections
               and rebinning)
      M = 1 for epoch 1 (launch - 3/21/96)
          2 for epoch 2 (3/21-4/14/96)
          3 for epoch 3 (4/14/96 to present)


The major improvement (relative to v2.1.2) is that the jumps predicted in the number of electrons produced at the L edge has been reduced. The v2.1.2 matrices produced large negative residuals near and below 4.8 keV, and often large positive residuals near 6 keV. An ad-hoc optimization of the size of the jump resulted in a 50% reduction relative to the theoretical values of Dias et al. (Phys Rev A, 48, 2887, 1993). Independently, Dias and collaborators made a more detailed comparison of their model with measurments of the apparent energy step across the edge (Journal of Applied Physics, Volume 82, Issue 6, September 15, 1997, p 2742) as measured in 3 different experiments. The Dias model predicts a larger edge than is measured by a factor of 1.3 to 2.5, so our ad-hoc adjustment gains some credence. Other improvements are that we now explicitly account for the electronic LLD cut-off at low channels, and we account for the mylar and propane thickness individually by detector.

Ron Remillard has summarized his experience with these matrices for fitting the Crab. Here is his text and tarred, gzipped, figures . This summary looks at each detector individually.

PCARMF v2.1.2

version 2.1 uses (by default) an energy to channel model that converts incident photon energy to number of electrons produced, and then uses a second order polynomial to relate number of electrons to channel (e2c_model=3). The number of electrons is estimated from the work of Dias et al. (Phys Rev A, 48, 2887, 1993 and private communication). This naturally introduces jumps in the energy to channel scale at the atomic absorbtion edges. The second order polynomial that relates incident energy to channel is not supported in this version. pcarmf v2.1.2 will only run with "e2c_model=3" Version 2.1.2 consistently handles the issue of adding the responses from the different layers. Now the sum of three matrices made for different layers of a particular detector is identical to a matrix made for the whole detector. This was handled inconsistently in previous versions.
Link removed. This contains an executable version of pcarmf (pcarmf), the parameter file with current values of the default settings (pcarmf.par), the energy to channel conversion coefficients (PCA.energy2chan.v2.1.2), and the source code and a Makefile to build pcarmf (note that users outside GSFC would need to ensure that the references to ftools libraries are correct for their installations). The parameters are slightly different than for version 2.1.2, but it is now possible to make matrices for all 3 epochs (the only exception is for detectors 3 and 4 in the second epoch; these detectors were not on much during this interval, and I currently suggest ignoring these detectors.)
Link removed. This file contains matrices with names pN_lM_eL.rsp. N runs from 0-4 and identifies the PCU. M runs from 1-4 and identifies the layer, with 4 being the propane layer. L runs from 1 -3 and identifies the epoch (1 is IOC through March 1996, 2 runs from 3-21-96 to 4-14-96, and 3 runs from 4-14-96 to the present) A matrix suitable for the entire PCU0 may be constructed by adding together the 3 PCU0 matrices for individual layers.
Examples of v2.1.2 performance are available.

PCARMF v2.1.1

version 2.1 uses (by default) an energy to channel model that converts incident photon energy to number of electrons produced, and then uses a second order polynomial to relate number of electrons to channel (e2c_model=3). The number of electrons is estimated from the work of Dias et al. (Phys Rev A, 48, 2887, 1993 and private communication). This naturally introduces jumps in the energy to channel scale at the atomic absorbtion edges. A second order polynomial that relates incident energy to channel is also available (e2c_model=2) though this model is inferior in the vicinity of the Xenon L absorption edges. version 2.1.1 consistently handles the issue of adding the responses from the different layers. Now the sum of three matrices made for different layers of a particular detector is identical to a matrix made for the whole detector. This was handled inconsistently in previous versions.
Link removed. This contains an executable version of pcarmf (pcarmf), the parameter file with current values of the default settings (pcarmf.par), the energy to channel conversion coefficients (PCA.energy2chan.v2.1.1), and the source code and a Makefile to build pcarmf (note that users outside GSFC would need to ensure that the references to ftools libraries are correct for their installations).
. Link removed. This file contains 15 matrices with names pNlrM_256.rsp. N runs from 0-4 and identifies the PCU. M runs from 1-3 and identifies the layer. A matrix suitable for the entire PCU0 may be constructed by adding together the 3 PCU0 matrices for individual layers.
Channel conversion coefficients and matrices for epoch 1 and 2 will be made available shortly.


PCARMF v2.02 version 2.02 is the version of PCARMF released with ftools 3.6.1. The default energy to channel conversion uses a second order polynomial that relates incident energy to channel number. Version 2.02 reuqires the use of XSPEC 9.02 or later and the use of matrix manipulation tools (gcorrmf, rbnrmf, xpcaarf, marfrmf, and addrmf) from ftools 3.6.1 or later. Use of pcarmf v2.02 and earlier versions of these tools risks inconsistent handling of the channel counting between the matrix generator and XSPEC. The fundamental issue is that the PCA pulse heights range from 0-255 while XSPEC counts spectral channels from 1-256. I believe that confusion on this isuue is behind us.
Link removed. (Source code is not supplied here; the curious should investigate the more recent v2.1.1 available above.)
Link removed. Matrices for epochs 1 and 2 may be requested from keith.jahoda@gsfc.nasa.gov or created from the executable above.


USAGE example
Here is a little script used to make a matrix. In this version I do not call xpcaarf, but rather assume that each detector has an area of 1400 cm2.


#!/bin/csh -f
#
# this script run from a directory below where the program, energy 2 channel
#     data, and parameter file live
#
set path = (~/pcarmf/v2.02 $path)
rehash
setenv SYSPFILES ~/pfiles/develop
setenv PFILES ~/pfiles/develop
cp ../pcarmf.par ~/pfiles/develop
cp ../PCA.energy2chan.v2.02 PCA.energy2chan.v2.02
which pcarmf

pcagainset phafile=brise1.pha gainfile=caldb mode=h
rddescr phafil=brise1.pha chanfil=chan.txt mode=h

foreach pcu (0 1 2 3 4)
    pcarmf pcuid=$pcu lld_code=63 date=960214 nofits=1 outfile=fred.rmf area_factor=1400.
    gcorrmf infile=fred.rmf shftfile=pcu$pcu"_shft.txt" outfile=sh$pcu.rsp
    rbnrmf infile=sh$pcu.rsp outfile=$pcu.rsp binfile=chan.txt mode=h clobber=yes
end

rm input2addrmf
ls ?.rsp | awk '{print $1, "1.0"}' > input2addrmf

addrmf listfile=input2addrmf rmffile=pca.rsp


exit



return to: RXTE GOF home page; About RXTE page; PCA page