SIS Energy-Channel Relationship (Gain and Offset)

Currently, the accuracy of gain calibration varies among the eight CCD chips on the two SIS detectors:

SIS0-chip1, the default chip for point sources, is believed to have the most accurate gain.
SIS1-chip3, also the default chip for point sources, is the next best, but its gain is probably 0.5 per cent too high.
The other, non-default chips may be off by as much as 1.5-2 per cent.

For more details, please consult the article by Dotani et al. in ASCANews no.3.

In addition, spurious but small (up to 30 eV) offsets may be present in the energy-channel relationship due to:

DFE (Dark Frame Error), and
RDD (Residual Dark Distribution) effects in the case of Faint mode data analyzed using the current version of the ftool faintdfe.

If your SIS spectra, even if extracted from the PI column, show features at energies where the SIS effective area curve has a large derivative (near the oxygen edge and gold M edge, for example), or if best-fitting line energies appear at the "wrong" energies, then this may be due to the inaccurate calibration of the energy-channel relationship - especially in the case of non-default chips.

You can explore this possibility using the gain fit command in xspec. Since the uncertainty in the slope of the gain is likely to be larger than the uncertainty in the offset, users should experiment first by adjusting the slope. For this, we recommend a two-stage process:

Use the xspec gain fit command to find the best value for the slope. This is straightforward:
1. Fit the spectrum using the nominal gain.
2. If applicable, fix the energies of line features at their expected values. For example, fix the energy of the fluorescent line from neutral iron at 6.4 keV.
3. At the xspec prompt type gain fit, after which you'll be prompted for the initial values of the slope and the offset.
4. Hit return for the slope (to make it a free parameter) and then 0,-1 to fix the offset at zero.
5. Next use the fit command to refit the data and derive the best-fitting slope which will be listed as the penultimate parameter. It should come out to be close to unity.
Repopulate the PI column in your SIS event file using the ftool sispi but with the hidden parameter gainnom to correct the gain. If, for example, you find the best-fitting gain to be 1.015, then the value of gainnom is 3.65*1.015, i.e., 3.705 (the default setting is 3.65 eV per channel). The corresponding command string is:
```
> sispi gainnom=3.705
```
After repopulating the PI column, re-extract the source and background spectra and repeat the fit.

If, after following this procedure, you still suspect that the gain is not correct, you can repeat the first step, but allowing the offset to vary rather than the slope. Note, however, that the present version of sispi cannot apply an offset correction to the energy-channel relationship.

Please send comments and questions to ascahelp@athena.gsfc.nasa.gov.

Koji Mukai and Charles Day, ASCA GOF

If you have any questions concerning ASCA, visit our Feedback form.

This file was last modified on Friday, 25-Feb-2000 14:58:56 EST