Although S+A may sound like a good way to obtain sharper images, in practice it suffers from a number of problems, making its use limited to bright compact sources, not faint extended structure. In order to benefit from S+A, the exposure times need to be shorter than the typical variations in seeing (ie as short as possible), for which the readout noise will dominate the background noise. And because there are a large number of readouts per stored frame, the summed readout noise will be quite large. The effective noise will be even larger if the number of transient hot pixels confuses the shift algorithm to occasionally shift to the hot pixel rather than the source, hence the limiting magnitude of the guide star needs to be lower than just a shorter exposure time might imply. As well, to obtain very good images, a magnified pixel scale is required, but there isn't enough room at GHRIL (in direct mode with the dichroic) to insert the pair of magnifying mirrors. S+A also requires an offset sky frame to be taken beforehand, which is then subtracted from all the object frames before they are shifted and added. And the number of exposures per integration must, for some peculiar reason, be a prime number.
You would only ever want to use S+A to sharpen your images, hence you need exposure times shorter than the seeing variation timescale. A reasonably bright star (brighter than 12 mag) is therefore needed for the S+A centring algorithm to determine the correct shifts. Because of the extra readout overhead with an ND_STARE, S+A is taken in STARE mode. This means a SKY frame must be taken before a S+A sequence, so it can be subtracted from the subsequent OBJECT frames within ALICE, in order remove most of the hot pixels (but alas not all - see § 3.4) and the fixed pattern of the bias pedestal and flat field, so that centroiding can be done efficiently. SKY subtraction also minimises the extra noise created by dithering the fixed pattern noise. So before running S+A, offset to a clear patch of sky, and take a sky frame of the same exposure as your object.
The sequence of commands to get S+A going is:
Once the OBJECT integration is finished, a normal (but hopefully better resolved) WHIRCAM image file will be created, plus a vector map of all the centroids determined by ALICE, in an image filename suffixed with `V'.