While the algorithms used for these extraction modes are explained in details in the Calnic C manual , it is important to understand the basic differences between them. These options allow to change the way that NICMOSlook combines the rows from the grism/spectrum image in order to construct the object's spectrum.
Note that NICMOSlook can be used to determine the size and orientation of an object, but that the user can still force the extraction to assume that the object is a point source (i.e. the cross dispersion of the spectrum is that of the PSF). On the other hand, if the size of an object has not been defined, the "Extended object" extraction will default back to to perform a point source extraction.
Since the spectrum in the grism/spectrum image contains less information as we move away from the center of the spectrum (in the cross dispersion direction), it is tempting to want to weight the various rows in a manner that takes into account their varying signal-to-noise ratios. When using a weighted extraction, NICMOSlook uses the knowledge of the cross dispersion shape of the spectrum (assumed for point source objects, or measured in the direct image for extended objects) to weight each row separately. When using an un-weighted extraction, NICMOSlook weights all extracted rows equally and uses its knowledge of the shape of the spectrum only to determine how many rows should be extracted from the grism/spectrum image (The current cut-off is set to a certain percentage of the central peak, as set by the parameter WEIGHTPERC).
The NICMOS 3 camera severely under-samples the PSF. In addition, the spectra are not exactly aligned along the rows. For that reason, a weighted extraction of a relatively compact source often results in spurious features when the spectrum crosses from one row of the image to the next one. This effect could only be modeled if both the exact position and shape of the object within one pixel and the pixel response function were known. Since this information is in general not available, NICMOSlook does not attempt to take this effect into account. Even if this information were available, only very little improvement of the signal-to-noise ratio could be expected by using a weighted extraction for such object.
