SIS observations were scheduled for July 3-6, 1994 (3 nights, spectroscopy) and August 6-8, 1994 (2 nights, deep B and I imaging). The instrumental configuration consisted of the B600 grism and a LORAL 2048
2048 CCD detector (``LORAL3''). Slit widths were 0and 0. The CCD detector was binned by 2 in both the spatial and dispersion directions for improved signal-to-noise ratios. The SIS instrumental configuration is described in Table 
, and the characteristics of the LORAL CCD are given in Table .
SIS observing involved a complex procedure which required up to 20 minutes of set-up overhead per field for an experienced observer. The first step was to rotate the telescope's Cassegrain bonnette to the position angle of the major axis of the primary target and to move the telescope to the target field. Once the telescope was in position, a direct, short exposure image was taken to determine the exact pixel coordinates of the primary target and the guide star. The guide probe was then moved to the guide star coordinates, and SIS guiding was initiated. The telescope position was then fine-tuned with a series of SIS-offsets such that all the target objects were centered on their respective slitlets. SIS-offsets moved the telescope while maintaining the guide probe locked on the guide star. The final telescope position was checked by taking a direct (no filter, no grism) image through the multi-object mask. Multi-object masks were created by laser cutting slitlets at the positions of the targets measured on a direct image.
The SIS observation log is shown in Table . Typical total integration time per galaxy was 7200 seconds. For each galaxy, the total exposure time was split between 2
3 exposures to identify and flag pixels hit by cosmic rays (see section ). In July, 3/4 of the first night was lost to very poor seeing conditions (FWHM > 1), and mediocre quality images were obtained without SIS fast guiding. Towards the end of the night, it was possible to take direct images for mask making. The first half of the second night was lost due to clouds, and a conflict between SIS guiding and the Telescope Control System (TCS). The telescope kept losing its position lock whenever SIS guiding was turned on.
The first multi-object spectroscopy exposure taken after solving the tracking problem revealed that something was terribly wrong with the mask alignment. There was a small rotation between the multi-object mask and the direct image used to create it. The positions of slitlets far from the center of the field were completely off their intended targets. The problem remained unsolved until after the second SIS run in August. It was traced back to lateral screw adjustments in the SIS mask holders which had never been adjusted since the commissioning of SIS. Each of the four SIS mask holders suffered from different amounts of rotation. Therefore, SIS had to be used in long-slit mode for the remainder of the July run and all of the August run, and the number of targets actually observed was considerably lower than expected. Three objects were successfully observed during the third night of the July run.
Weather conditions during the August run were good with a typical seeing of 0. Only two hours were lost when the CCD warmed up unexpectedly. Unfortunately, the seeing during these two hours went down to 0! Two objects were observed on the first night. Observations of a third object were ruined by high CCD temperature. Four objects were observed on the second night.