20pt1) Be very careful about running more than one SAS task at a time in the same directory. This can have unforeseen consequences which may not be obvious, and will often produce erroneous results. It is our strong recommendation that this not be done, ever.

20pt2) In this document when examples of task calls are provided they are broken into multiple lines for clarity. When actually invoking a task all parameters should be on one line (or the lines should be ended with the proper continuation character).

20pt3) Observations with small filtered exposures ($\sim5$ ks or less) are likely to be problematic due to the limited number of events in the unexposed corners of the detectors; decent statistics are needed for accurate background modeling. Exposures in the $\sim5-10$ ks range can also be problematic, particularly for the pn, with the QPB being over- or under-subtracted. When the QPB is under-subtracted the fitting of the SP component can correct for this effect, but in an uncertain manner with unknown systematics.

20pt4) Fitting the SP component when the residual contamination is relatively small can produce unreasonable values for the spectral index, especially for the pn. The fitted indicies can run away to large values, either positive or negative. In general, reasonable indicies lie between 0.15 and 1.5 when fitting a single power law or 0.1 and 2.5 for a broken power law. Note that when the fitted value for the SP normalization drops below a few times $10^{-3}$ (for the full FOV) the values are effectively zero. There is no practical difference between normalizations of $10^{-4}$ and $10^{-30}$.

20pt5) Double-pixel events for the pn contribute an unacceptably high background below 0.4 keV and occasionally can affect data up to $\sim0.6$ keV. Because of this we recommend using only single-pixel (PATTERN == 0) events when data less than 0.7 keV are of interest. On the other hand, excluding double-pixel events (PATTERN $<$= 4) from the study of higher energies forgoes a significant part of the signal. We therefore recommend the following complication: process the data twice, creating two pn spectra with single- and double-pixel events. The spectra can then be simultaneously fit including $0.3-2.0$ keV data from the single-pixel spectrum and $1.0-11.0$ keV data from the double-pixel spectrum.

20pt6) Multiple pn segments may exist for your observation. While emchain will process all MOS segments within an obsid, epchain will only process the first segment, unless you specifically tell it which segment to process. You will have to determine the existence of the other segments and process them separately. This is discussed in ยง5.7.

20pt7) If you update the CCF directory contents, cifbuild must be rerun.

20pt8a) If cifbuild or odfingest have any problems, you should delete the ccf.cif file and/or the *.SAS file in the odf directory and rerun cifbuild and odfingest after correcting the issue. A “corrupt” ccf.cif or *.SAS file will allow SAS to try to do things if it's there, and fail with often difficult to decipher error messages. (Our helpdesk guru notes that this issue is the most common source of problems.)

20pt8b) The environment variable SAS_CCF points to the ccf.cif that all tasks will use. Similarly, the *.SAS file in the directory pointed to by SAS_ODF. If you move on to another observation, SAS_CCF and SAS_ODF must be changed, or else you will not get the right results. We suggest doing only one obsid per window.

20pt9) ESAS is undergoing major changes as it is being made more SAS compliant. If you have used ESAS in the past, please check the change logs for a first look at where your analysis methods/scripts may need to be changed.