1) The location of the zero order image (hint, use the RCP command to locate).
2) The angle the dispersed spectrum makes with the X-axis (typically -ROLL_ANG-0.07 for LETG, -ROLL_ANG-4.725 for HEG and -ROLL_ANG+5.235 for HEG).
3) How wide the spectrum is on the detector? All widths represent the maximum distance from the centerline and hence are half widths. Thus all pixels less than the S_width from the centerline will be in the source spectra. All less that B_out but greater than B_in will be included in the background spectra. Events between the S_width and B_in are considered to be in no man's land and are not included in either spectra.
4) The number of pixels per channel. This is typically 1.0. Since the HRC oversamples the PSF, the LETG/HRC convention is to use 2 pixels per spectral channel.
5) Do you wish to sum +/- orders? Typically one wants to sum counts from the plus and minus orders. This puts the zero order in the maximum channel. If you don't sum, the spectrum will contain twice the number of channels and the zero order will be in the center.
If a graphics device is active, then PLT will display the extracted spectrum. Since the 0 order image is very bright, you may need replot on a LOG scale to really see the spectrum. When you quit out of PLT you will be asked "Prefix for PHA file name:" if you don't want to create an XSPEC FITS file, then enter a blank line. For a non- blank entry, then a FITS file with that name will be created.
Counts with bad width:
after the event list has been read.