|
|
|
Sample filtering results are shown below for a spectrum of Capella (obsid1248, 85 ksec). The first plot shows the extracted and background-subtracted X-ray spectrum (using the bowtie-shaped extraction region), the second plot shows the accompanying background "spectrum", and the third plot shows the fraction of background that remains after the various filters are applied.
Gaps around -53 and +62 Å arise from the gaps between detector segments. The number of background counts per wavelength bin (which were extracted from rectangular regions on either side of the X-ray extraction region) has been normalized to estimate the background level within the extracted X-ray spectrum.
The effect of simply removing all events with pha=255 is also shown for comparison. This filtering, which can be safely applied to all HRC-S data (even at wavelengths shorter than 6 Å, and to observations without any grating) cleans up a significant number of invalid events (which create much of the sawtooth character in the background plot) and removes roughly 25% of the background with essentially zero X-ray signal loss.
| BG-subtracted spectrum | Filtered background | Relative background fraction |
|---|---|---|
 |
 |
 |
Background reductions of roughly 50% can be obtained for non-dispersed images (and for 0th order in grating data) as follows: Extract a small region around the source, small enough so that relative background contamination will be negligible. The pi spectrum of this data subset will then correspond to the pure X-ray source spectrum. Next, select a range of pi values that keep whatever fraction of events you wish (e.g., pi=40:160 might keep 98% of the events). One way to do this is to use dmstat to determine the mean and standard deviation for pi and then use dmlist with varying pi filter ranges to see what fraction of events are retained. Every source will be different, so you'll have to determine what range to use. Then apply that filtering range using dmcopy to your entire data set to reduce the background level. It's just like filtering on energy in ACIS data, only the resolution is much worse. Not clear? Email me.
Gory details regarding how this filtering method was derived (somewhat out
of date and not very organized) are described
elsewhere.
A second higher-resolution version
of the gain map used to convert pha to pi, and further study
of the relationship between in-flight and laboratory gains and between
photon wavelength and pi values, is just getting started.
We hope to have an updated gain map and filter regions by the end
of the year which may squeeze out another 5% or so of the background.
A better characterization of the pulse-height distribution would
permit better estimates of the effect on removal of higher-order photons,
but would require a great deal more effort.
Last modified: 08/22/07
 |
The Chandra X-Ray
Center (CXC) is operated for NASA by the Smithsonian Astrophysical Observatory. 60 Garden Street, Cambridge, MA 02138 USA. Email: cxcweb@head.cfa.harvard.edu Smithsonian Institution, Copyright © 1998-2004. All rights reserved. |