When the target is extended or there are several targets spread over an arcminute or more, the full ACIS array is probably the best choice and is the "standard" mode for HETGS observations. Another commonly used subarray is to use rows 1-774 with a SIM_Z offset of -3 mm. This subarray is often used when the source is point-like and when the low energy portion of the spectrum (with highest resolution) is important to the observer. The minimum (optimal) frametime is reduced to 2.5 s. One may compute the number of rows in the subarray required to maintain energy E at the end of the HEG spectrum or 0.5E at the end of the MEG spectrum and the corresponding SIM_Z offset that starts the subarray at row 1 and centers 0th order in the subarray:
Y_subarray = min(1024, 2*Y_bg + 32 + 389/[E/keV] )
SIM_Z = (Y_subarray/2 - 512) * 0.024
where Y_bg is the desired background size on either side of the subarray (in pixels), the 32 comes from an allowance for dither, and SIM_Z is the offset required to center the source in the subarray (in mm). The SIM_Z is negative for such subarrays.
For E = 0.65 keV, which is the minimum energy achievable for a -20" Y offset, and for Y_bg = 70 pix (about 10 x the spectrum half-width) gives 770 rows and -3.05 mm. For a heavily obscured but otherwise bright source where there is very little or no flux below 1 keV, one might consider reducing the background region to 50 pix on either side of the spectrum, giving a subarray size of 521 rows and the corresponding SIM_Z offset would be -6.04 mm (so that the subarray starts at row 1). In this case, the frame time can be reduced to 1.7 s if all 6 ACIS-S chips are read out. (Chip S0 would not have useful data but turning it off would not affect the frametime in this case.)
See the HETGS GO Info web page for more issues to consider when planning a Chandra HETGS observation.