|
|
|
The new HRC-S gain map and LETG pulse-height filter replace older versions from 2001, and were officially released in CALDB 4.2.0 (December 2009). As shown in the plots below, over half of Level 2 background events can be removed at most wavelengths with a loss of ~1.25% of X-ray events. Among the differences from the old gain map and PI filter are:
Electrical field-fringing effects on the ends of the three HRC-S segments create high background emission. The current badpix map doesn't cut off all of this high-background region, and users may wish to apply a little more spatial filtering, such as rawy=:16383,17070:32200,32980:. In any case, spectral features near the plate gaps should be treated with suspicion. [This issue was addressed with an updated BADPIX map in May 2017.]
The pulse-height filter removes only 1.25% of 1st order X-ray events, but an increasing fraction of higher-order events. The mean of the pulse-height distribution increases slowly with photon energy, such that a factor of two difference in energy corresponds to a shift in the mean of ~8%. The mean of 8th order will therefore be about 25% higher than 1st order. As an example, the filter removes 1.25% of 1st order at 160 Å and about 11% of 8th order (λ=20 Å, mλ=160 Å). Extra filtering of higher orders will have negligible effect for most analyses, but should be considered when deliberately studying wavelength ranges with very heavy higher order contamination. Users may instead wish to apply 'Flat Filtering' (see below), which removes a negligible fraction of X-ray events for all orders.
Great effort was expended to make the new gain map and filter generally applicable to all observations, even those with large offset pointings. Gain calibration near the HRC-S plate gaps and around the default aimpoint, however, is not quite as good as elsewhere. For observations with large offsets, users may wish to compare results using the standard spectroscopy filter versus those obtained using less restrictive 'Flat Filtering' (see below), although we do not expect any statistically significant differences.
In some cases, observers may wish to apply filtering that removes
a constant fraction of the background across the entire HRC-S.
Examples include analyses of extended or off-axis sources
(with or without a grating),
and spectra with heavy higher-order contamination.
The following command
$ dmcopy "evt2.fits[pi=0:300]" evt2_filtered opt=all
will remove about 20% of background events with negligible
X-ray event loss.
Because of 'burn in' around the default aimpoint (Yoffset,Zoffset = 0,0), the gain in that region varies significantly on very small scales and is, for various reasons, difficult to calibrate accurately. Given the usually high S/N for point sources, the small improvement that would be gained by reducing the background, and the potential for nonuniform background removal because of gain calibration uncertainties, PI filtering is generally not recommended when the objective is to analyze non-grating or 0th-order images near the default aimpoint.
 |
 |
| Comparison of old and new filters, applied to data from 2000 and 2008. Background fractions are relative to Level 2 data that include events with PHA=255. The old PI filter became less effective over time because it did not account for decreasing gain. Less background is removed at short wavelengths because the filtering threshold must be higher to avoid excluding X-ray events. The raw BG rate (not shown) is higher in 2008 than 2000 because of the solar cycle, but a higher fraction of BG is removed in 2008 because the BG pulse-height distribution has relatively more high-channel events. | The 2008 LETG/HRC-S background rate with and without filtering, using the standard 'bowtie' spectral extraction region. The filtered rate will decline by roughly half as we approach Solar Maximum in 2012/13. The right-hand axis units refer to the ~0.07-Å line width (~1.4 times the FWHM) of the LETGS. |