Spectral fit of newly re-processed ground calibraion data

Goro Sato, Derek Hullinger
28 Dec 2004

We generate new pha files using batgse2dph and batbinevt, because batgse2dph calculates the energy scale exactly the same way bateconvert does, so the results will be closer to the actual data that will be coming through the ground processing pipeline. The pha files that we have already fit were generated by a different set of tools (full_rebin and sum_spectra) that assigned the energy in a slightly different way. Then, we re-fit these new pha files for coarse grid runs and far field runs so that we can update the mutau function parameters (stored in swbparams***.fits in CALDB) to be used by batdrmgen.

Derek's note:
One thing that is interesting is that the peaks in the new summed pha files are systematically off by a tiny amount (slightly too high at lower energies and slightly too low at higher energies). I noticed while I was investigating the energy scaling problem that if voltage is expressed as a quadratic function of channel, the assigned energy has a greater systematic difference from the true energy than when channel is expressed as a quadratic function of voltage. bateconvert (and batgse2dph) uses a voltage=f(channel) equation to assign energy, whereas full_rebin used a channel=f(voltage) equation. So the full_rebin-generated files have a slightly better energy assignment. It would probably take a fairly major software change to alter the batbinevt algorithm, so I'm inclined to just accept the systematic discrepancy in the counts spectrum (which is less than 0.5 keV) and correct for it in batdrmgen instead (in the form of new fit parameters).



Results

I notice that our new procedure gives about 30% smaller error bars for each PHA spectrum (I don't know why). Therefore, resultant reduced chi-squared is about twice larger than before. But, the fit is still good enough to use the fitted parameters. It is also worth pointing out that the parameter "gain_index" becomes less than 1.0 while it was more than 1.0. This is due to the change of energy conversion program. The results are shown below for both coarse grid runs and far field runs. The details are available in the linked pages.

Coarse Grid runs :

Details

Far Field runs :

Details

Return to main BAT response page