IMPORTANT: Problems and Corrections to RXTE PCA Background

[RXTE GOF <xtenews@bigbang.gsfc.nasa.gov>]

Summary:

Problems were discovered in the SAA history file that result in errors
in the background rate estimation. A new file that corrects these
problems is available.

In addition, a bug was discovered in the FTOOL PCABACKEST for faint
models only, in Epoch 5c. The parameter maxmodels=2000 (instead of the
default value maxmodels=600) must be set in order to return a complete
faint background model. This can be done by the command:

 myprompt> pset pcabackest maxmodels=2000

immediately before calling pcabackest.

The SAA history problems and FTOOLS bug, their likely impact on
analysis of PCA data, and remedies are described in detail on the RXTE
PCA Digest Page (What's New).

Details:

Recently, several problems with the data in the RXTE SAA history file
have been identified by the PCA team. This file, which is based on
HEXTE particle monitor rates, is used in the PCA background model to
estimate the time-varying contribution of the PCA background component
due to the activation caused by passage through the South Atlantic
Anomaly (SAA).

Three types of problems were identified:

1. Missing data ("gaps"): time intervals, on the order of a day or
more, where no data are present in the SAA history file. The typical
gap is a day, but the longest gap is 15 days (2007Apr 30 through
2007May14).

2. Duplicated data: SAA passes are included in the file multiple
times.  About 13 days are affected by this problem;

3. Dropped points: about half the time, the first SAA pass in a new
day is missing from the file.

All of the problems have been corrected in a new version of the file,
now available for use. The file can be downloaded from the link on the
PCA Digest Page:

 http://heasarc.gsfc.nasa.gov/docs/xte/pca_news.html

where there is also a link to a PCA Team page with a detailed report
about the SAA history file problems and the PCABACKEST bug.

Missing SAA passes cause underestimation of the activation component
in the background model, whereas duplicated SAA passes cause
overestimation.

While investigating the SAA history file problems, it was discovered
that the maxmodels parameter in the tool PCABACKEST is set to a value
too small to notice the activation components in the faint model for
Epoch 5C.  This will result in underestimation of the background for
all observations using the faint background model during Epoch 5C
(2004Jan01-present).  This can be fixed by setting the maxmodels
parameter to 2000 when calling PCABACKEST. A patched version of this
FTOOL will be available shortly.

Impacts:

While the exact size of the error in the background estimates for a
particular observation can only be determined by reanalyzing the data,
we provide the following guidance.

The SAA gaps and duplicates can cause background errors up to 20%
(bright model) and 10% (faint model) on the days they occurred.  We
find that 60% of bright observations tested show less than a 2%
difference in background rate due to this effect. The amount of the
effect is a function of the model (bright versus faint), the energy
band, and time.

For the faint model, the biggest impact is a daily variation in the
2-5 KeV band (<10%) in the 5 orbits per day which have SAAs. The
typical error is half this value. There is less than a 2% effect above
5 keV. Note that observations from 2004Jan01 and later which use the
faint background model are affected by both the SAA history file
problem and the PCABACKEST bug.

For the bright model, the biggest impact is a 20-30% underestimate in
the calculated background rate for all energy bands, within 50 minutes
of an SAA, due to the gaps in the SAA history file.  The impact on
observations far from an SAA is small to negligible. The bright model
is not impacted by the bug in PCABACKEST.

Remedies:

New file: A corrected SAA history file has been generated by the PCA
team.  This file is based on all HEXTE particle monitor data in the
archive.  Future releases of the SAA history file will be based upon
this corrected file, and checks are in place to verify that new data
is free from these problems.

Users of the faint model MUST set maxmodels=2000 when invoking the
tool in order for the activation terms in the model to be applied.

Advice: users are advised to get the new SAA history file and
reanalyze their data using PCABACKEST maxmodels=2000 to look for the
presence of the effect of background mis-estimation. Note that many
observations occur far from passage of RXTE through the SAA. For these
observations, virtually no effect of a bad SAA file is
present. However, it is a complex process to identify which
observations may have been affected by the problems in the SAA
file. For this reason we urge RXTE users with data from 2004Jan01 and
later to reanalyze their data with the new SAA history file and
maxmodels=2000.

--
A service of the RXTE GOF, NASA/GSFC

Please do not reply to this email.
Questions or comments should go to:
http://rxte.gsfc.nasa.gov/cgi-bin/Feedback