The VLA is a versatile instrument. It is capable of observing at six wavelengths in four different configurations. It is not sufficient to decide to do a survey, one must also decide which survey to do. The choice of a frequency was an easy one to make. The field of view of the VLA scales linearly with wavelength (inversely with frequency.) A survey at 6 cm would require times as many observations as a survey at 20 cm covering the same area of sky. The same logic would suggest that the VLA's longest wavelength, 90 cm, would be even better than 20 cm. However, the surface density of bright sources is much higher at 90 cm so that a single snapshot of data is insufficient to achieve the high dynamic range required for reaching the theoretical sensitivity of the VLA. Furthermore, the 90 cm receivers are more than an order of magnitude less sensitive than the 20 cm receivers.
The relatively high resolution of the B configuration ( arcsec) can also be justified despite the extra difficulty that high-resolution imaging entails. There are two primary drawbacks to going to higher resolution. A B-configuration image has times as many resolution elements per unit area as a D-configuration image, hence times as much mass storage is needed to archive images and the computational time to produce each image increases by a factor of . In addition, VLA images will degrade if the bandwidth chosen is too wide (bandwidth smearing). The farther apart the antennas, the smaller is the acceptable bandwidth. In B configuration, this constraint forces the observations to be made in spectral line mode, resulting in the loss of 50%of the available bandpass compared to continuum mode.
However, the scientific advantages are worth the operational disadvantages. In particular, the high resolution results in better positional accuracy for all sources found in the survey and, for a significant fraction of sources, results in morphological classification based on a source's brightness distribution. Much of the power of the radio survey will come from the association of individual radio sources with optical counterparts. But unambiguous association demands a radio error box small enough to exclude random coincidences, where ``small enough'' depends on both the density of potential optical counterparts and the fraction of optical sources that have radio emission. Approximately 50%of the radio sources in the FIRST survey will have optical counterparts brighter than 23rd magnitude. At this level, sub-arcsecond positions are necessary for reliable identifications.
The morphology of the brightness distribution of a radio source is also important for making a correct optical identification. Many sources are complex; for example, radio triples are common. In such cases, the optical counterpart will be located at the position of a weak central component, far from the brighter radio lobes. Without high resolution, misidentification is likely. The morphology is also important for classifying the nature of the radio emission (e.g., distinguishing between FR I and FR II radio sources.)