You will want to work with a moderately bright star, for efficiency (6-7 mag is about right). Put the star on the cross hair in panoramic mode. (The stage should previously have been set to the nominal position of the IRS slit. Put the IRS in imaging mode and take an image of the slit: mirror; go (assuming a standard grating is on sidea). Measure the slit center (in pixels). Then open the aperture and image the star in the field: ap to open_hole; go. Now center the star on the position in the field that corresponds to the slit center. You should have "follow on" so that the x-y stage will follow the movements of the telescope keeping the star on the cross hair. Since you will not be guiding during this, there may be some drift and some iterations will be necessary.
NOTE: One can image the field using the Oth order of the standard gratings or the mirror (but not the cross-dispersed grating). Which should be used is a question of how far the grating has to move and imaging requirements. The mirror has poor image quality but may be closer to the normal observing position of the grating.
Once centered the night assistant should zpoint the stage to establish the "middle" position as the current IRS slit center.
The same procedure should now be repeated for the corner of the chip used for guiding. This is more critical but also easier, because the guider should be on during the procedure.
Having established carefully the position of the slit center in imaging mode, execute the Tcl command: zrecenter <xposition(pix)><yposition(pix)>. <rot. angle>. This command initializes the parameters necessary for the recentering and offsetting routines, which take into account the angle of rotation of the rotator.