Dear Chandra user,
the correction for time-dependent ACIS gain has been updated. The
updates described below include a bug fix for the S3 chip, more time
coverage, correction for S1, and new software.
T-gain correction should be implemented within acis_process_events in
the next CIAO release. In the meantime, the equivalent software and
calibration data are available on the Chandra software exchange page,
http://cxc.harvard.edu/cont-soft/software/corr_tgain.1.0.html
Our general approach to deriving the time-dependent gain correction is
described in http://cxc.harvard.edu/contrib/alexey/tgain/tgain.html.
***************** BUG FIX **************************************************
The previous version of t-gain files was produced using the results of
fitting the External Cal. Source data with RMFs computed by CIAO's
mkrmf. MKRMF has a bug which results in shifting the response depending on
the user-supplied energy grid. If the grid starts above ~0.3 keV, it is
believed that the RMF is correct; if the grid starts near 0.1 keV (which is
the default in some popular spectral extraction scripts), the shift is
introduced.
Because mkrmf has been used inconsistently in our previous analysis, t-gain
data for the S3 chip ONLY had an artificial ~8-eV correction.
To eliminate any dependence on the software bugs, I have used a new approach
for computing t-gains. The locations of the cal. source lines are measured
without any spectral fits, and t-gain correction is computed relative to the
data collected over February-April of 2000.
The implications for the users are the following:
OLD VERSION OF TGAIN RELEASED IN THE SUMMER OF 2003:
I0-I3, S2 chips: t-gains were correct. The users should not notice any
systematic shifts.
S3: t-gain data had a wrong shift by ~8eV. The users should notice this in
the spectral fits but the shift is different depending on how RMFs were
computed.
- if the spectrum was extracted in the PHA channels binned by 2, and
the energy grid started from 0.1 keV, the fit should be good because
of the double error.
- in all other combinations of channel and energy grids, the shifts of
~4-14 eV can be present.
THIS RELEASE OF TGAIN:
t-gain data does not rely on the mkrmf and should be correct. If you see a
systematic energy shift in your fits, it is related to the mkrmf bug or the
astrophysics or the 0-epoch energy scale calibration.
To avoid most problems, it is recommended to give MKRMF the energy grid with
the minimum energy above 0.3 keV until the bug is fixed.
The difference between the previous and new releases is minimal for the
ACIS-I and S2 chips. Although the correction is slightly more accurate now,
you don't need to reprocess the data which were already corrected using
tgain files from the previous release.
***************** MORE TIME AND CCD COVERAGE *********************************
t-gain files in this release cover the time period through the end of summer
of 2003, and contain the correction for the S1 chip.
***************** NEW SOFTWARE ***********************************************
The package contains a new program, apply_gain, which makes the correction
process more straightforward. In particular, you can recompute PI and ENERGY
values on a single pass (i.e. no need to run acis_process_events
afterwards). Please see README for details.
This archive was generated by hypermail 2b29 : Tue Feb 12 2013 - 01:00:11 EST